CN111458676B - Direction-of-arrival estimation method and device based on cascaded neural network - Google Patents
Direction-of-arrival estimation method and device based on cascaded neural network Download PDFInfo
- Publication number
- CN111458676B CN111458676B CN202010148361.6A CN202010148361A CN111458676B CN 111458676 B CN111458676 B CN 111458676B CN 202010148361 A CN202010148361 A CN 202010148361A CN 111458676 B CN111458676 B CN 111458676B
- Authority
- CN
- China
- Prior art keywords
- signal
- noise ratio
- sample
- network
- arrival
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
- G01S3/02—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using radio waves
- G01S3/14—Systems for determining direction or deviation from predetermined direction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The embodiment of the invention provides a method and a device for estimating a direction of arrival based on a cascaded neural network, which are used for determining a covariance matrix of received signals based on the received signals from a plurality of signal sources; performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector; converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors; inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr. Can be applied to a wide signal-to-noise ratio range.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a method and a device for estimating a direction of arrival based on a cascaded neural network.
Background
DOA (Direction of Arrival) Estimation is an important research topic in the wireless communication field, and aims to obtain the incident Direction of a signal when the signal reaches an antenna.
In the existing DOA estimation method based on the neural network, a covariance matrix is constructed based on received signals, the covariance matrix comprises all signal DOA information, elements in an upper triangular matrix of the covariance matrix of the received signals are input into the neural network which is constructed in advance and trained, and a signal DOA estimation result, namely a signal incidence direction angle, is output. However, the DOA estimation method has certain disadvantages, which are embodied as the following two points: 1) the method is not suitable for scenes with large signal-to-noise ratio range, and if the network training is carried out by using noise-free signal data, the estimation performance of the network on noisy signals is poor; if noisy signal data is used for network training, the network can obtain better estimation performance under the condition of low signal-to-noise ratio, but the estimation performance under the condition of high signal-to-noise ratio is not as good as that of the traditional DOA estimation algorithm. I.e. less adaptive to noise. 2) The DOA estimation can be carried out only for the signal transmitted by a single information source, and the DOA estimation method cannot be applied to the scenes of a plurality of information sources.
Therefore, a method suitable for a wide signal-to-noise ratio range and capable of simultaneously performing DOA estimation on signals transmitted by a plurality of information sources is needed.
Disclosure of Invention
The embodiment of the invention aims to provide a method and a device for estimating the direction of arrival based on a cascaded neural network, so as to realize the estimation of the direction of arrival of signals sent by a plurality of information sources in a wide signal-to-noise ratio range. The specific technical scheme is as follows:
in order to achieve the above object, an embodiment of the present invention provides a method for estimating a direction of arrival based on a cascaded neural network, where the method includes:
determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector;
converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
inputting the eigenvalue vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, and the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is the eigenvalue vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is the covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio, and the high signal-to-noise ratio estimation sub-network is activated; and the output result of the signal-to-noise ratio classification network activates the low signal-to-noise ratio estimation sub-network for low signal-to-noise ratio.
Optionally, the performing eigenvalue decomposition on the covariance matrix to obtain a first eigenvalue vector includes:
decomposing the covariance matrix R based on the following formula to obtain an eigenvalue vector sigma:
wherein, UsRepresenting a signal subspace, UnRepresenting a noise subspace in whichIs a real number, σ, related to the signal power of the kth signal sourcenRepresenting the noise power; lambdanRepresenting a noise power diagonal matrix, R having a vector of eigenvalues of
Optionally, the snr classification network is trained according to a first sample data set, and the generating step of the first sample data set includes:
generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of high signal-to-noise ratio and low signal-to-noise ratio; the first sample signal comprises a high signal-to-noise ratio signal and a low signal-to-noise ratio signal;
determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector;
identifying the vector of sample eigenvalues and the signal-to-noise ratio of the first sample signal as the first sample dataset;
the high signal-to-noise ratio estimation subnetwork is trained on a second sample data set, the generating of which comprises:
generating a sample high signal-to-noise ratio signal and a first sample direction-of-arrival angle corresponding to the sample high signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the restriction of the direction-of-arrival;
determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors;
taking the first sample direction of arrival angle and the first sample covariance information vector as the second sample data set;
the low signal-to-noise ratio estimation sub-network is trained on a third set of sample data, the generating of the third set of sample data comprising:
generating a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the restriction of the direction-of-arrival;
determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors;
taking the second sample direction of arrival angle and the second sample covariance information vector as the third sample data set.
Optionally, the high snr estimation subnetwork is trained according to the following steps;
acquiring a preset neural network and the second sample data set;
inputting the first sample covariance information vector into the preset neural network to obtain a direction of arrival angle output by the current neural network;
determining a loss value based on the direction of arrival angle output by the current neural network and the first sample direction of arrival angle;
and judging whether the neural network is converged or not based on the loss value, if so, determining the current neural network as a high signal-to-noise ratio estimation sub-network, and if not, returning to the step of inputting the first sample covariance information vector into the preset neural network to obtain the direction of arrival angle.
In order to achieve the above object, an embodiment of the present invention further provides a direction of arrival estimation apparatus based on a cascaded neural network, where the apparatus includes:
a determining module for determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
the decomposition module is used for decomposing the eigenvalue of the covariance matrix to obtain an eigenvalue vector;
the conversion module is used for converting the elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
the estimation module is used for inputting the eigenvalue vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, and the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is the eigenvalue vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is the covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio, and the high signal-to-noise ratio estimation sub-network is activated; and the output result of the signal-to-noise ratio classification network activates the low signal-to-noise ratio estimation sub-network for low signal-to-noise ratio.
Optionally, the decomposition module is specifically configured to:
decomposing the covariance matrix R based on the following formula to obtain a first eigenvector sigma:
decomposing the covariance matrix R based on the following formula to obtain an eigenvalue vector sigma:
wherein, UsRepresenting a signal subspace, UnRepresenting a noise subspace in whichIs a real number, σ, related to the signal power of the kth signal sourcenRepresenting the noise power; lambdanRepresenting a noise power diagonal matrix, R having a vector of eigenvalues of
Optionally, the snr classification network is trained on a first sample data set, the high snr estimation subnetwork is trained on a second sample data set, and the low snr estimation subnetwork is trained on a third sample data set, the apparatus further comprising a first generation module, a second generation module, a third generation module,
the first generation module is used for generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of a high signal-to-noise ratio and a low signal-to-noise ratio; the first sample signal comprises a high signal-to-noise ratio signal and a low signal-to-noise ratio signal;
determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector;
identifying the vector of sample eigenvalues and the signal-to-noise ratio of the first sample signal as the first sample dataset;
the second generation module is used for generating a sample high signal-to-noise ratio signal and a first sample direction-of-arrival angle corresponding to the sample high signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival;
determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors;
taking the first sample direction of arrival angle and the first sample covariance information vector as the second sample data set;
the third generating module is configured to generate a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival;
determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors;
taking the second sample direction of arrival angle and the second sample covariance information vector as the third sample data set.
Optionally, the apparatus further comprises a training module, wherein the training module is configured to:
acquiring a preset neural network and the second sample data set;
inputting the first sample covariance information vector into the preset neural network to obtain a direction of arrival angle output by the current neural network;
determining a loss value based on the direction of arrival angle output by the current neural network and the first sample direction of arrival angle;
and judging whether the neural network is converged or not based on the loss value, if so, determining the current neural network as a high signal-to-noise ratio estimation sub-network, and if not, returning to the step of inputting the first sample covariance information vector into the preset neural network to obtain the direction of arrival angle.
In order to achieve the above object, an embodiment of the present invention further provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus;
a memory for storing a computer program;
and the processor is used for realizing any method step when executing the program stored in the memory.
To achieve the above object, an embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, and the computer program, when executed by a processor, implements any of the above method steps.
Therefore, by adopting the method and the device for estimating the direction of arrival based on the cascaded neural network, provided by the embodiment of the invention, the covariance matrix of the received signals is determined based on the received signals from a plurality of signal sources; performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector; converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors; inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr. Therefore, compared with the traditional method for estimating the direction of arrival by adopting a single neural network, the method can improve the accuracy of the direction of arrival estimation, thereby being suitable for a wide signal-to-noise ratio range.
Of course, not all of the advantages described above need to be achieved at the same time in the practice of any one product or method of the invention.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a method for estimating a direction of arrival based on a cascaded neural network according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a cascaded neural network according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of generating a first sample data set according to an embodiment of the present invention;
fig. 4 is a schematic flowchart of generating a second sample data set according to an embodiment of the present invention;
fig. 5 is a schematic flowchart of generating a third sample data set according to an embodiment of the present invention;
FIG. 6 is a schematic flow chart of training a sub-network for high SNR estimation according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a direction of arrival estimation apparatus based on a cascaded neural network according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to solve the technical problem that the conventional direction of arrival estimation method cannot be applied to a large-range signal-to-noise ratio, the embodiment of the invention provides a direction of arrival estimation method and device based on a cascaded neural network, an electronic device and a computer-readable storage medium.
For ease of understanding, an application scenario of the embodiment of the present invention is described first.
The method for estimating the direction of arrival based on the cascaded neural network provided by the embodiment of the invention can be applied to a receiving end of a wireless communication system. The receiving end can receive signals sent by a plurality of signal sources at the same time, the incident angle of the signals sent by each signal source is usually different when the signals reach the receiving end antenna array, and the receiving end calculates the incident direction angle of the signals when the signals reach the antenna array according to the received signals, namely the direction of arrival estimation.
Referring to fig. 1, fig. 1 is a schematic flowchart of a method for estimating a direction of arrival based on a cascaded neural network according to an embodiment of the present invention, where the method may include the following steps:
s101: based on the received signals from the plurality of signal sources, a covariance matrix of the received signals is determined.
In the embodiment of the invention, K signal sources are arranged, and the signal from the kth signal source received at the time t is recorded as sk(t), wherein K is 1, 2. These signals are directed at the ULA (Uniform Linear Array) at the receiving end at different angles. The received signal at the receiving end can be expressed as:
x(t)=As(t)+n(t)
wherein s (t) ═ s1(t),s2(t),...,sK(t)]TIs a signal vector, n (t) represents statistical independent additive white gaussian noise, A represents an array manifold matrix, and x (t) represents a receiving signal of a receiving end.
Those skilled in the art will readily appreciate that the DOA estimation problem of a signal can be viewed as a mapping problem, with the objective of establishing a received signal x (t) and a direction of arrival angle θ1,θ2,θk,…θK]In which θ iskI.e. representing the k signalDirection of arrival angle of the source signal when it arrives at the receiving end.
In the embodiment of the invention, a receiving end can calculate a covariance matrix based on a received signal x (t), the covariance matrix comprises signal-to-noise ratio information of the received signal and direction-of-arrival information of the signal, if the direction-of-arrival is required to be directly obtained from the covariance matrix, a large amount of matrix inversion operation is involved, and the calculation complexity is high. Therefore, the neural network can be adopted to realize the mapping relation between the data vector of the received signal and the direction of arrival angle so as to rapidly solve the direction of arrival angle.
In the embodiment of the present invention, the covariance matrix of the received signal may be calculated based on the following formula:
wherein S represents a preset number of sampling signals.
S102: and carrying out eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector.
In the embodiment of the present invention, after determining the covariance matrix of the received signal, in order to extract information related to the signal-to-noise ratio of the signal from the covariance matrix, eigenvalue decomposition may be performed on the covariance matrix to obtain an eigenvalue vector, and the eigenvalue vector may be used as a basis for determining the signal-to-noise ratio level of the signal.
S103: and converting the elements in the covariance matrix into normalized real vectors to obtain covariance information vectors.
In the embodiment of the present invention, since the covariance matrix is a conjugate symmetric matrix, the upper triangular region of the covariance matrix may include the direction of arrival information of all received signals. However, since the elements in the covariance matrix are complex numbers, the elements in the upper triangular region of the covariance matrix can be real-valued, and used as a covariance information vector, which includes data information in the covariance matrix and can be used as an input vector of the neural network.
Specifically, let r denote the vectorized upper triangular matrix of the covariance matrix, and its construction method is as follows:
r=[R1,2,R1,3,…,R1,N,R2,3,…,R2,N…,RN-1,N]
wherein R isu,vThe elements in the u-th row and v-th column of the covariance matrix R are shown. r is a complex vector that cannot be used directly as input to the neural network, and is therefore converted to a normalized real vectorAs covariance information vectors, i.e.
Where Real (·) represents the operation of taking the Real part of the complex value, and imag (r) represents the operation of taking the imaginary part of the complex value.
In the embodiment of the present invention, the execution sequence of step S102 and step S103 is not limited.
S104: inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr.
In order to enable the neural network to be suitable for a wide signal-to-noise ratio range, in the embodiment of the present invention, a cascaded neural network structure is constructed, as shown in fig. 2, the cascaded neural network includes a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network includes a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network, an input of the signal-to-noise ratio classification network is an eigenvalue vector, and an input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector. The output of the signal-to-noise ratio classification network determines to activate a high signal-to-noise ratio estimation sub-network or a low signal-to-noise ratio estimation sub-network, and specifically, if the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio, the high signal-to-noise ratio estimation sub-network is activated; and if the output result of the signal-to-noise ratio classification network is low signal-to-noise ratio, activating a low signal-to-noise ratio estimation sub-network.
For example, the receiving end inputs the feature vector and the covariance information vector into a pre-trained cascade neural network, and then the feature vector is input into a signal-to-noise ratio classification network, wherein the signal-to-noise ratio classification network is a single hidden layer multiple input single output neural network, an output value is-1 or +1, -1 represents a low signal-to-noise ratio, and +1 represents a high signal-to-noise ratio. If the output of the signal-to-noise ratio classification network is-1, the low signal-to-noise ratio estimation sub-network is activated, namely the covariance information vector is input into the low signal-to-noise ratio estimation sub-network, and the low signal-to-noise ratio estimation sub-network can output the estimation result of the direction of arrival of the signal; on the contrary, if the output of the snr classification network is +1, the snr estimation sub-network is activated, i.e., the covariance information vector is input to the snr estimation sub-network, and the snr estimation sub-network can output the estimation result of the direction of arrival of the signal.
In addition, in another embodiment of the present invention, the output result of the snr classification network may be divided into a plurality of levels, each level corresponding to a snr range, and accordingly, the direction of arrival estimation network is divided into a plurality of sub-networks, each snr range corresponding to a sub-network in the active direction of arrival estimation network.
For example, the output result of the snr classification network is three levels, i.e., a low snr, a medium snr, and a high snr, and the direction-of-arrival estimation network may include a low snr estimation sub-network, a medium snr estimation sub-network, and a high snr estimation sub-network. And activating the low signal-to-noise ratio estimation sub-network if the output result of the signal-to-noise ratio classification network is the low signal-to-noise ratio, activating the middle signal-to-noise ratio estimation sub-network if the output result of the signal-to-noise ratio classification network is the middle signal-to-noise ratio, and activating the high signal-to-noise ratio estimation sub-network if the output result of the signal-to-noise ratio classification network is the high signal-to-noise ratio. The grade and range of signal-to-noise ratio division can be set according to actual requirements, and the embodiment of the invention does not limit the grade and range.
Therefore, by adopting the estimation method of the direction of arrival based on the cascaded neural network provided by the embodiment of the invention, the covariance matrix of the received signals is determined based on the received signals from a plurality of signal sources; performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector; converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors; inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr. Therefore, compared with the traditional method for estimating the direction of arrival by adopting a single neural network, the method can improve the accuracy of the direction of arrival estimation, thereby being suitable for a wide signal-to-noise ratio range.
In one embodiment of the present invention, step S102 may include:
decomposing the covariance matrix R based on the following formula to obtain a first eigenvector sigma:
wherein U issRepresenting a signal subspace, UnRepresenting a noise subspace, the two subspaces being orthogonal to each other; whereinIs a real number, σ, related to the signal power of the kth signal sourcenRepresenting the noise power, ΛnIs a noise power diagonal matrix whose non-zero elements are equal toThus, the eigenvalue vector of R is represented asIt is a real number vector and contains signal-to-noise ratio information, which can be used as an effective basis for determining the signal-to-noise ratio level.
In one embodiment of the invention, the snr classification network, the high snr estimation sub-network, and the low snr estimation sub-network in the cascaded neural network may each be trained independently.
The snr classification network may be trained according to a first sample data set, see fig. 3, and the generating step of the first sample data set may include:
step a 1: generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of high signal-to-noise ratio and low signal-to-noise ratio; the first sample signal includes a high signal-to-noise ratio signal and a low signal-to-noise ratio signal.
Classification thresholds of high and low signal-to-noise ratios can be preset, and then a large number of high signal-to-noise ratio signals and low signal-to-noise ratio signals are generated and are used as first sample signals, wherein the signal-to-noise ratio of the high signal-to-noise ratio signals is marked as the high signal-to-noise ratio, and the signal-to-noise ratio of the low signal-to-noise ratio signals is marked as the low signal-to-noise ratio.
Step a 2: and determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector.
Similar to step S102, a first sample covariance matrix of the first sample signal is determined and subjected to eigenvalue decomposition, and a large number of sample eigenvalue vectors can be obtained.
Step a 3: the vector of sample eigenvalues, and the signal-to-noise ratio of the first sample signal are identified as a first sample data set.
And carrying out eigenvalue decomposition on the sample covariance matrix to obtain a sample eigenvalue vector, and using the sample eigenvalue vector and a corresponding high signal-to-noise ratio or low signal-to-noise ratio identifier as a first sample data set for training a signal-to-noise ratio classification network.
The high snr estimation subnetwork may be trained on a second set of sample data, see fig. 4, which may be generated by:
step b 1: based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraints of the direction of arrival, generating a sample high signal-to-noise ratio signal and a first sample direction of arrival angle corresponding to the sample high signal-to-noise ratio signal.
In the embodiment of the invention, the sine value range of the direction of arrival angle is [ -1,1], a large number of sample high signal-to-noise ratio signals can be generated under the classification threshold values of high signal-to-noise ratio and low signal-to-noise ratio and the constraint conditions of the direction of arrival, and the sample beam direction angle corresponding to each sample high signal-to-noise ratio signal is determined.
Step b 2: and determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors.
Similar to step S103, a second sample covariance matrix of the sample high snr signal is determined, and elements in the second sample covariance matrix are converted into normalized real vectors, so that a large amount of first sample covariance information vectors can be obtained.
Step b 3: taking the first sample direction of arrival angle and the first sample covariance information vector as a second sample data set;
the first sample covariance information vector and the corresponding sample beam direction angle under the high signal-to-noise ratio can be used as a second sample data set for training the high signal-to-noise ratio estimation subnetwork.
The low snr estimation sub-network may be trained on a third sample data set, see fig. 5, which may be generated by:
step c 1: and generating a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival.
In the embodiment of the invention, the sine value range of the direction of arrival angle is [ -1,1], a large number of sample low signal-to-noise ratio signals can be generated under the classification threshold values of high signal-to-noise ratio and low signal-to-noise ratio and the constraint conditions of the direction of arrival, and the sample beam direction angle corresponding to each sample low signal-to-noise ratio signal is determined.
Step c 2: and determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors.
Similar to step S103, a third sample covariance matrix of the sample low snr signal is determined, and elements in the third sample covariance matrix are converted into normalized real vectors, so that a large amount of second sample covariance information vectors can be obtained.
Step c 3: and taking the direction of arrival angle of the second sample and the covariance information vector of the second sample as a third sample data set.
The second sample covariance information vector and the corresponding sample beam direction angle under the low signal-to-noise ratio can be used as a third sample data set for training the low signal-to-noise ratio estimation subnetwork.
In the embodiment of the invention, the training data of the signal-to-noise ratio classification network, the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network are different, and the training processes are mutually independent. The training process of the network is described below by taking the high snr estimation subnetwork as an example.
Specifically, referring to fig. 6, the high snr estimation sub-network can be trained as follows:
s601: acquiring a preset neural network and a second sample data set;
the preset neural network represents a high signal-to-noise ratio estimation sub-network to be trained, and all parameters in the network are initialization parameters.
S602: inputting the second sample feature vector into a preset neural network to obtain a direction of arrival angle output by the current neural network;
s603: determining a loss value based on the direction of arrival angle output by the current neural network and the direction of arrival angle of the sample contained in the second sample data set;
in the embodiment of the present invention, the loss value is obtained by using, but not limited to, MSE (Mean Squared Error) formula as the loss function.
S604: and judging whether the neural network converges or not based on the loss value, if so, executing S605, otherwise, returning to the step S602.
S605: the current neural network is determined as a high signal-to-noise ratio estimation subnetwork.
Based on the same inventive concept, according to the embodiment of the method for estimating a direction of arrival based on a cascaded neural network, the embodiment of the present invention further provides a device for estimating a direction of arrival based on a cascaded neural network, as shown in fig. 7, the device may include the following modules:
a determining module 701 for determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
a decomposition module 702, configured to perform eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector;
a conversion module 703, configured to convert elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
an estimating module 704, configured to input the eigenvalue vector and the covariance information vector into a pre-trained cascaded neural network, so as to obtain direction of arrival angles of signals for each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr.
In an embodiment of the present invention, the decomposition module 702 may be specifically configured to:
decomposing the covariance matrix R based on the following formula to obtain an eigenvalue vector sigma:
wherein, UsRepresenting a signal subspace, UnRepresenting a noise subspace in whichIs a real number, σ, related to the signal power of the kth signal sourcenRepresenting the noise power; lambdanRepresenting a noise power diagonal matrix, R having a vector of eigenvalues of
In one embodiment of the invention, the snr classification network is trained on a first sample data set, the high snr estimation subnetwork is trained on a second sample data set, and the low snr estimation subnetwork is trained on a third sample data set, the apparatus further comprising a first generation module, a second generation module, a third generation module,
the first generation module is used for generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of high signal-to-noise ratio and low signal-to-noise ratio; the first sample signal comprises a high signal-to-noise ratio signal and a low signal-to-noise ratio signal;
determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector;
identifying the sample eigenvalue vector and the signal-to-noise ratio of the first sample signal as a first sample dataset;
the second generation module is used for generating a sample high signal-to-noise ratio signal and a first sample direction-of-arrival angle corresponding to the sample high signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the restriction of the direction-of-arrival;
determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors;
taking the first sample direction of arrival angle and the first sample covariance information vector as a second sample data set;
the third generation module is used for generating a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival;
determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors;
and taking the direction of arrival angle of the second sample and the covariance information vector of the second sample as a third sample data set.
In an embodiment of the present invention, on the basis of the apparatus shown in fig. 7, a training module may be further included, where the training module is configured to:
acquiring a preset neural network and a second sample data set;
inputting the first sample covariance information vector into a preset neural network to obtain a direction of arrival angle output by the current neural network;
determining a loss value based on the direction of arrival angle output by the current neural network and the first sample direction of arrival angle;
and judging whether the neural network is converged or not based on the loss value, if so, determining the current neural network as a high signal-to-noise ratio estimation sub-network, otherwise, returning to the step of inputting the first sample covariance information vector into the preset neural network to obtain the direction of arrival angle.
By adopting the wave arrival direction estimation device based on the cascade neural network provided by the embodiment of the invention, the covariance matrix of the received signals is determined based on the received signals from a plurality of signal sources; performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector; converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors; inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr. Therefore, compared with the traditional method for estimating the direction of arrival by adopting a single neural network, the method can improve the accuracy of the direction of arrival estimation, thereby being suitable for a wide signal-to-noise ratio range.
Based on the same inventive concept, according to the above-mentioned embodiment of the method for estimating the direction of arrival based on the cascaded neural network, an embodiment of the present invention further provides an electronic device, as shown in fig. 8, which includes a processor 801, a communication interface 802, a memory 803, and a communication bus 804, wherein the processor 801, the communication interface 802, and the memory 803 complete mutual communication through the communication bus 804,
a memory 803 for storing a computer program;
the processor 801 is configured to implement the following steps when executing the program stored in the memory 803:
determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector;
converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr.
The communication bus mentioned in the electronic device may be a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface is used for communication between the electronic equipment and other equipment.
The Memory may include a RAM (Random Access Memory) or an NVM (Non-Volatile Memory), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also a DSP (Digital Signal Processing), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component.
By adopting the electronic equipment provided by the embodiment of the invention, the covariance matrix of the received signals is determined based on the received signals from a plurality of signal sources; performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector; converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors; inputting the characteristic value vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, wherein the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is a characteristic value vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is a covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio activation high signal-to-noise ratio estimation sub-network; the output of the snr classification network activates the low snr estimation sub-network for low snr. Therefore, compared with the traditional method for estimating the direction of arrival by adopting a single neural network, the method can improve the accuracy of the direction of arrival estimation, thereby being suitable for a wide signal-to-noise ratio range.
Based on the same inventive concept, according to the above-mentioned embodiments of the cascaded neural network-based direction of arrival estimation method, in yet another embodiment provided by the present invention, a computer-readable storage medium is further provided, in which a computer program is stored, and the computer program is executed by a processor to implement the above-mentioned method steps shown in fig. 1 to 6.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The procedures or functions according to the embodiments of the invention are brought about in whole or in part when the computer program instructions are loaded and executed on a computer. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wirelessly (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium such as a Solid State Disk (SSD), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the embodiments of the apparatus for estimating direction of arrival, the electronic device and the computer-readable storage medium based on the cascaded neural network, since they are substantially similar to the embodiments of the method for estimating direction of arrival based on the cascaded neural network, the description is simple, and the relevant points can be found in the partial description of the embodiments of the method for estimating direction of arrival based on the cascaded neural network.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (8)
1. A method for estimating a direction of arrival based on a cascaded neural network, the method comprising:
determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
performing eigenvalue decomposition on the covariance matrix to obtain an eigenvalue vector;
converting elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
inputting the eigenvalue vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, and the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is the eigenvalue vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is the covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio, and the high signal-to-noise ratio estimation sub-network is activated; the output result of the signal-to-noise ratio classification network is low signal-to-noise ratio, and the low signal-to-noise ratio estimation sub-network is activated;
the snr classification network is trained from a first sample data set, the generating of the first sample data set comprising:
generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of high signal-to-noise ratio and low signal-to-noise ratio; the first sample signal comprises a high signal-to-noise ratio signal and a low signal-to-noise ratio signal;
determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector;
identifying the vector of sample eigenvalues and the signal-to-noise ratio of the first sample signal as the first sample dataset;
the high signal-to-noise ratio estimation subnetwork is trained on a second sample data set, the generating of which comprises:
generating a sample high signal-to-noise ratio signal and a first sample direction-of-arrival angle corresponding to the sample high signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the restriction of the direction-of-arrival;
determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors;
taking the first sample direction of arrival angle and the first sample covariance information vector as the second sample data set;
the low signal-to-noise ratio estimation sub-network is trained on a third set of sample data, the generating of the third set of sample data comprising:
generating a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the restriction of the direction-of-arrival;
determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors;
taking the second sample direction of arrival angle and the second sample covariance information vector as the third sample data set.
2. The method of claim 1, wherein performing eigenvalue decomposition on the covariance matrix to obtain a first eigenvalue vector comprises:
decomposing the covariance matrix R based on the following formula to obtain an eigenvalue vector sigma:
3. The method of claim 1, wherein the high signal-to-noise ratio estimation sub-network is trained by;
acquiring a preset neural network and the second sample data set;
inputting the first sample covariance information vector into the preset neural network to obtain a direction of arrival angle output by the current neural network;
determining a loss value based on the direction of arrival angle output by the current neural network and the first sample direction of arrival angle;
and judging whether the neural network is converged or not based on the loss value, if so, determining the current neural network as a high signal-to-noise ratio estimation sub-network, and if not, returning to the step of inputting the first sample covariance information vector into the preset neural network to obtain the direction of arrival angle.
4. A cascaded neural network-based direction-of-arrival estimation apparatus, the apparatus comprising:
a determining module for determining a covariance matrix of received signals based on the received signals from the plurality of signal sources;
the decomposition module is used for decomposing the eigenvalue of the covariance matrix to obtain an eigenvalue vector;
the conversion module is used for converting the elements in the covariance matrix into normalized real vectors to obtain covariance information vectors;
the estimation module is used for inputting the eigenvalue vector and the covariance information vector into a pre-trained cascade neural network to obtain the direction of arrival angle of the signal aiming at each signal source; the cascade neural network comprises a signal-to-noise ratio classification network and a direction-of-arrival estimation network, and the direction-of-arrival estimation network comprises a high signal-to-noise ratio estimation sub-network and a low signal-to-noise ratio estimation sub-network; the input of the signal-to-noise ratio classification network is the eigenvalue vector, and the input of the high signal-to-noise ratio estimation sub-network and the low signal-to-noise ratio estimation sub-network is the covariance information vector; the output result of the signal-to-noise ratio classification network is high signal-to-noise ratio, and the high signal-to-noise ratio estimation sub-network is activated; the output result of the signal-to-noise ratio classification network is low signal-to-noise ratio, and the low signal-to-noise ratio estimation sub-network is activated;
the SNR classification network is trained on a first sample data set, the high SNR estimation sub-network is trained on a second sample data set, the low SNR estimation sub-network is trained on a third sample data set, the apparatus further comprises a first generation module, a second generation module, a third generation module,
the first generation module is used for generating a first sample signal and a signal-to-noise ratio identifier of the first sample signal based on a preset classification threshold value of a high signal-to-noise ratio and a low signal-to-noise ratio; the first sample signal comprises a high signal-to-noise ratio signal and a low signal-to-noise ratio signal;
determining a first sample covariance matrix of the first sample signal, and performing eigenvalue decomposition on the first sample covariance matrix to obtain a sample eigenvalue vector;
identifying the vector of sample eigenvalues and the signal-to-noise ratio of the first sample signal as the first sample dataset;
the second generation module is used for generating a sample high signal-to-noise ratio signal and a first sample direction-of-arrival angle corresponding to the sample high signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival;
determining a second sample covariance matrix of the sample high signal-to-noise ratio signal, and converting elements in the second sample covariance matrix into normalized real vectors to obtain first sample covariance information vectors;
taking the first sample direction of arrival angle and the first sample covariance information vector as the second sample data set;
the third generating module is configured to generate a sample low signal-to-noise ratio signal and a second sample direction-of-arrival angle corresponding to the sample low signal-to-noise ratio signal based on the classification threshold values of the high signal-to-noise ratio and the low signal-to-noise ratio and the constraint of the direction-of-arrival;
determining a third sample covariance matrix of the sample low signal-to-noise ratio signal, and converting elements in the third sample covariance matrix into normalized real vectors to obtain second sample covariance information vectors;
taking the second sample direction of arrival angle and the second sample covariance information vector as the third sample data set.
5. The apparatus of claim 4, wherein the decomposition module is specifically configured to:
decomposing the covariance matrix R based on the following formula to obtain an eigenvalue vector sigma:
6. The apparatus of claim 4, further comprising a training module to:
acquiring a preset neural network and the second sample data set;
inputting the first sample covariance information vector into the preset neural network to obtain a direction of arrival angle output by the current neural network;
determining a loss value based on the direction of arrival angle output by the current neural network and the first sample direction of arrival angle;
and judging whether the neural network is converged or not based on the loss value, if so, determining the current neural network as a high signal-to-noise ratio estimation sub-network, and if not, returning to the step of inputting the first sample covariance information vector into the preset neural network to obtain the direction of arrival angle.
7. An electronic device is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for realizing mutual communication by the memory through the communication bus;
a memory for storing a computer program;
a processor for implementing the method steps of any of claims 1 to 3 when executing a program stored in the memory.
8. A computer-readable storage medium, characterized in that a computer program is stored in the computer-readable storage medium, which computer program, when being executed by a processor, carries out the method steps of any one of the claims 1-3.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010148361.6A CN111458676B (en) | 2020-03-05 | 2020-03-05 | Direction-of-arrival estimation method and device based on cascaded neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010148361.6A CN111458676B (en) | 2020-03-05 | 2020-03-05 | Direction-of-arrival estimation method and device based on cascaded neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111458676A CN111458676A (en) | 2020-07-28 |
CN111458676B true CN111458676B (en) | 2022-03-29 |
Family
ID=71685571
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010148361.6A Active CN111458676B (en) | 2020-03-05 | 2020-03-05 | Direction-of-arrival estimation method and device based on cascaded neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111458676B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114501353B (en) * | 2020-10-23 | 2024-01-05 | 维沃移动通信有限公司 | Communication information sending and receiving method and communication equipment |
CN112946564B (en) * | 2021-04-12 | 2024-02-02 | 西北大学 | DOA estimation method and device based on DNN (digital optical network) beam space and computer storage medium |
CN116318470B (en) * | 2023-01-09 | 2024-05-10 | 中国电子科技集团公司第三十六研究所 | Method and device for estimating communication interference signal power under non-Gaussian noise |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5262789A (en) * | 1992-04-30 | 1993-11-16 | General Electric Company | Source identification system for closely separated spatial sources |
CN104616059A (en) * | 2015-01-29 | 2015-05-13 | 无锡职业技术学院 | DOA (Direction of Arrival) estimation method based on quantum-behaved particle swarm |
CN106156451A (en) * | 2015-03-16 | 2016-11-23 | 江南大学 | A kind of based on the Mutual coupling technology improving quantum particle swarm |
CN108037520A (en) * | 2017-12-27 | 2018-05-15 | 中国人民解放军战略支援部队信息工程大学 | Direct deviations modification method based on neutral net under the conditions of array amplitude phase error |
CN109709510A (en) * | 2018-12-24 | 2019-05-03 | 贵州航天计量测试技术研究所 | A kind of estimation method and system of coherent 2-d direction finding |
CN110531313A (en) * | 2019-08-30 | 2019-12-03 | 西安交通大学 | A kind of near-field signals source localization method based on deep neural network regression model |
CN110690930A (en) * | 2019-09-27 | 2020-01-14 | 清华大学 | Information source number detection method and device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7714778B2 (en) * | 1997-08-20 | 2010-05-11 | Tracbeam Llc | Wireless location gateway and applications therefor |
-
2020
- 2020-03-05 CN CN202010148361.6A patent/CN111458676B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5262789A (en) * | 1992-04-30 | 1993-11-16 | General Electric Company | Source identification system for closely separated spatial sources |
CN104616059A (en) * | 2015-01-29 | 2015-05-13 | 无锡职业技术学院 | DOA (Direction of Arrival) estimation method based on quantum-behaved particle swarm |
CN106156451A (en) * | 2015-03-16 | 2016-11-23 | 江南大学 | A kind of based on the Mutual coupling technology improving quantum particle swarm |
CN108037520A (en) * | 2017-12-27 | 2018-05-15 | 中国人民解放军战略支援部队信息工程大学 | Direct deviations modification method based on neutral net under the conditions of array amplitude phase error |
CN109709510A (en) * | 2018-12-24 | 2019-05-03 | 贵州航天计量测试技术研究所 | A kind of estimation method and system of coherent 2-d direction finding |
CN110531313A (en) * | 2019-08-30 | 2019-12-03 | 西安交通大学 | A kind of near-field signals source localization method based on deep neural network regression model |
CN110690930A (en) * | 2019-09-27 | 2020-01-14 | 清华大学 | Information source number detection method and device |
Non-Patent Citations (2)
Title |
---|
《Neural network-Based smart antenna for multiple source tracking》;Zooghby A H El et.al;《Transactions Antennas and Propagation》;20001231;第48卷(第5期);第768-776页 * |
《基于径向基神经网络的波达方向估计算法》;何宏 等;《徐州工程学院学报》;20120930;第27卷(第3期);第6-10页 * |
Also Published As
Publication number | Publication date |
---|---|
CN111458676A (en) | 2020-07-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111458676B (en) | Direction-of-arrival estimation method and device based on cascaded neural network | |
Huang et al. | New designs on MVDR robust adaptive beamforming based on optimal steering vector estimation | |
Flam et al. | On MMSE estimation: A linear model under Gaussian mixture statistics | |
CN107450047A (en) | Compressed sensing DOA estimation method based on unknown mutual coupling information under nested battle array | |
CN111856402A (en) | Signal processing method and device, storage medium, and electronic device | |
Cotter | Multiple snapshot matching pursuit for direction of arrival (DOA) estimation | |
Du et al. | Adaptive beamforming algorithm for coprime array based on interference and noise covariance matrix reconstruction | |
Coutino et al. | Direction of arrival estimation based on information geometry | |
Sarangi et al. | Super-resolution with sparse arrays: A nonasymptotic analysis of spatiotemporal trade-offs | |
Askari et al. | Robust beamforming in circular arrays using phase‐mode transformation | |
Fesl et al. | On the mean square error optimal estimator in one-bit quantized systems | |
CN113051739A (en) | Robustness self-adaptive processing method based on sparse constraint | |
CN116736342A (en) | GNSS spoofing detection method and system based on integrated CNN | |
Ebrahimi et al. | Generalised two stage cumulants‐based MUSIC algorithm for passive mixed sources localisation | |
Blanco et al. | Sparse covariance fitting for direction of arrival estimation | |
Ramezanpour et al. | DNN‐based interference mitigation beamformer | |
Changgan et al. | An improved forward/backward spatial smoothing root-MUSIC algorithm based on signal decorrelation | |
CN114184999B (en) | Method for processing generated model of cross-coupling small-aperture array | |
WO2020215911A1 (en) | Method and apparatus for classification of mixed signals, and electronic device | |
CN114047481A (en) | Robust adaptive beam forming method based on subspace orthogonality | |
Chen et al. | Optimal joint detection and estimation in linear models | |
Jian et al. | A adaptive beamforming design in low sample number conditions based on diagonal loading algorithm | |
D'ortenzio et al. | Fixed-point iterative computation of Gaussian barycenters for some dissimilarity measures | |
Karim et al. | A new improved algorithm for low complexity wideband angle of arrival estimation | |
D'Ortenzio et al. | Fixed-point iterations for several dissimilarity measure barycenters in the gaussian case |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |