CN115580340A - Method for inhibiting same frequency interference of digital satellite based on neural network - Google Patents
Method for inhibiting same frequency interference of digital satellite based on neural network Download PDFInfo
- Publication number
- CN115580340A CN115580340A CN202211253844.8A CN202211253844A CN115580340A CN 115580340 A CN115580340 A CN 115580340A CN 202211253844 A CN202211253844 A CN 202211253844A CN 115580340 A CN115580340 A CN 115580340A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- neuron
- neural network
- neurons
- 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.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 79
- 238000000034 method Methods 0.000 title claims abstract description 29
- 230000002401 inhibitory effect Effects 0.000 title claims abstract description 6
- 210000002569 neuron Anatomy 0.000 claims abstract description 221
- 230000001537 neural effect Effects 0.000 claims description 37
- 238000000605 extraction Methods 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000010287 polarization Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
- H04B7/18513—Transmission in a satellite or space-based system
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/20—Adaptations for transmission via a GHz frequency band, e.g. via satellite
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Astronomy & Astrophysics (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Aviation & Aerospace Engineering (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for inhibiting same frequency interference of a digital satellite based on a neural network. The invention comprises a neural network pre-training part and an online operation part. The neural network pre-training part firstly adds random white Gaussian noise to the generated standard signal to generate a random co-frequency interference signal and obtain a receiving analog signal; and extracting real part data and imaginary part data of the received analog signal at each moment, sending the real part data and the imaginary part data into the input layer neural network to obtain the output of the output layer neuron, judging and iterating to obtain a group of neural network coefficients, and storing the neural network coefficients in a register of the chip. The online operation part firstly inputs IQ signals into a neural network to obtain the output of output layer neurons, and carries out judgment and iteration, wherein the output of the neural network at each moment is the output after the same frequency interference is inhibited. The method of the invention presets the coefficient into the chip, resists the influence of same frequency interference according to the coefficient trained in advance when the chip works normally, and can also adapt to the current channel state in real time and in a self-adaptive manner.
Description
Technical Field
The invention belongs to the technical field of digital satellite television broadcasting, and particularly relates to a method for inhibiting same frequency interference by a digital satellite based on a neural network, which is used for resisting the same frequency interference caused by pseudo-orthogonality of satellite broadcasting signals in the same frequency band in a polarization direction.
Background
In the technical field of digital satellite television broadcasting, due to the fact that polarization directions of electromagnetic waves are not well aligned, some low-quality satellite receiving antennas cause a phenomenon that two original electromagnetic wave signals orthogonal in the same frequency band interfere with each other, which is called Co-Channel interference (CCI), as shown in fig. 1. There are many possible causes of co-channel interference, for example, different satellites occupy the same frequency band, and co-channel interference occurs. The quality of a target signal received by a receiver is seriously influenced by the same frequency interference problem, and the stronger the same frequency interference signal is, the lower the signal-to-interference-and-noise ratio of the received target signal is. In the field of digital satellite broadcasting technology, when the co-channel interference is strong to a certain extent, a mosaic phenomenon occurs in a received television program, and even a target signal cannot be normally received.
The prior art for solving the problem of co-channel interference generally includes the following methods:
and (3) reconstruction offset: based on some known information, the interfering signal is reconstructed and cancelled using a subtractor. This is the idea used in patent CN104253775a, for example.
The transmitting side avoids the channel interference: when the transmitter senses that the frequency band has the interference signal, the current frequency band is intelligently skipped, and the signal is not transmitted on the interference frequency band. This is the method used by the preamble puncturing technique of WiFi7, patent CN106569235a, for example.
And (3) constructing optimal time-frequency joint interleaving under narrow-band interference through special coding: this is the case, for example, in patent CN 108737029B.
Decision feedback equalization techniques: the channel equalizer coefficient is adjusted by using part of the known training data to achieve the purpose of canceling the interference signal. This is the case, for example, in patent CN 101340206B.
But none of the above methods are suitable for digital satellite broadcast applications. Since the receiver has no reference signal, the reconstruction cancellation method cannot be applied; for both methods 2 and 3, the transmitting end needs to be adjusted, and the related methods cannot be used. For part of satellite digital television broadcasting signals, under the condition that available training data does not exist, the problem of co-channel interference cannot be solved well by applying a decision feedback method.
Patent CN106569235B proposes to use an external rf switch to adjust the antenna position to suppress co-channel interference, but for a civil digital satellite set-top box, extra cost is undoubtedly added, and it is difficult to apply to the civil market. The patent CN104253775B is dedicated to a TD-SCMDA communication system receiver, and it adopts a method of determining and reconstructing cancellation by interference signals to suppress co-channel interference, but a digital satellite receiver has only one receiving antenna, so that there is no signal that can be provided for reference, which results in that the interference signals cannot be reconstructed well, and the reconstruction cancellation method described in CN104253775a cannot be used. Patent CN104049242B performs staggered timing processing on the originating signal, and additionally superimposes interference and a sliding window to realize asynchronous synchronous interference, but in the application scenario of digital satellite tv broadcasting, the originating signal cannot be adjusted, so that the originating signal cannot be applied to the digital satellite tv broadcasting scenario. The patent CN111245454B uses multiple transmission signals and hot standby redundancy to resist the burst co-channel interference signal, and obviously, in the civil digital broadcast television system, the multiple transmitters and multiple receivers redundancy are difficult to be widely applied in the market.
Currently, an Artificial Intelligence (AI) technology is widely applied to various fields, and a core technology of the AI technology is an Artificial Neural Network (ANN). The ANN consists of a plurality of neuronal structures, which are shown in fig. 2. The ANN can be used for fitting a certain function, and under the condition that the coefficient of the ANN is properly trained, the same-frequency interference signal can be virtually reconstructed by using the ANN, so that the target signal is directly recovered. Based on such a concept, the invention described herein is proposed.
Disclosure of Invention
The invention aims to provide a method for inhibiting same frequency interference by a digital satellite based on a neural network, which is completely different from the existing realization path for solving the problem of same frequency interference, and trains a group of network coefficients to be preset in ANN in a chip according to a same frequency interference scene by using the plasticity of an artificial neural network. And in the chip operation stage, dynamically updating the network coefficient according to the real-time demodulation data.
The invention comprises a neural network pre-training part and an online operation part.
The neural network pre-training part specifically comprises the following steps:
generating a standard signal with the length of L according to a digital satellite television broadcast signal, wherein the standard signal is one of QPSK, 8PSK, 16APSK, 32APSK, 64APSK, 128APSK and 256 APSK; random white Gaussian noise is added at the theoretical threshold of the standard signal; taking the sum of standard signal and random white Gaussian noise as tag signal S 1 ＝[s 1，1 ，s 1，2 ，...，s 1，L ]，s 1，l L =1,2, which is the L-th tag signal 1，l Is a plurality of numbers.
Step (2) generating co-channel interference signal S with random length L 2 ，S 2 ＝[s 2，1 ，s 2，2 ，...，s 2，L ]，s 2，l L =1,2, which is the L co-channel interference signal 2，l Is a plurality;
will S 2 Is added to S 1 To obtain a received analog signal S 3 ＝S 1 +S 2 ，S 3 ＝[s 1，1 +s 2，1 ，s 1，2 +s 2，2 ，..，s 1，L +s 2，L ]。
Step (3) adding S 3 The real part and the imaginary part of the signal are respectively sent into two groups asIn the shift register of the information extraction layer; at t 1 Time of day extraction S 3 1 to Kth data at t 2 Time of day extraction S 3 Sequentially operating the data from the 2 nd to the K +1 th according to the time sequence, and extracting the data at the L-K moments, wherein K & lt L; the data extracted at each time comprises real part data I x ＝[x I，1 ，x I，2 ，...，x I，K ]And imaginary data Q x ＝[x Q，1 ，x Q，2 ，...，x Q，K ]。
Extracting data on the shift register and sending the data into an input layer neural network, calculating the output F (x) = x + tanh (x) of each neuron in 2K neurons of an input layer, wherein the tanh (.) is a hyperbolic tangent function, and the input variable x corresponds to the data I extracted at each moment x ＝[x I，1 ，x I，2 ，...，x I，K ]And Q x ＝[x Q，1 ，x Q，2 ，...，x Q，K ]；
The output of the input layer neural network comprises an I-way output F IL0 And Q path output F QL0 ：
Step (5) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the output F of the neural unit of the I path IL1 And neural element output of the Q-way F QL1 ；F IL1 ＝[f IL1，1 ，f IL1，2 ，...，f IL1，J ]，f IL1，j Represents the jth neuron output of the first layer I of the hidden layer, F QL1 ＝[f QL1，1 ，f QL1，2 ，...，f QL1，J ]，f QL1，j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2., J and J are the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m，k，j For the coefficient values of the j-th neuron in the first layer of the hidden layer connected to the k-th neuron of the input layer, the initial values are all 1.
And (6) calculating the output of a second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the second layer of neurons is equal to the number J of the first layer of neurons, and the output of the second layer of neurons comprises the output F of the neural unit of the path I IL2 Neural element output F of sum Q QL2 ；F IL2 ＝[f IL2，1 ，f IL2，2 ，...，f IL2，J ]，f IL2，j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2，1 ，f QL2，2 ，...，f QL2，J ]，f QL2，j Represents the J-th neuron output of the Q-way of the second layer of the hidden layer, J =1,2., J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m，k，j The coefficient values of the jth neuron in the second layer of the hidden layer and the kth neuron in the first layer of the hidden layer are connected, and the initial values are all 1.
Step (7) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the third layer of neurons is equal to the number J of the first layer of neurons, the output of the third layer of neurons comprises the output F of the neural unit of the path I IL3 And neural element output of the Q-way F QL3 ；F IL3 ＝[f IL3，1 ，f IL3，2 ，...，f IL3，J ]，f IL3，j Represents the jth neuron output of the third layer I of the hidden layer, F QL3 ＝[f QL3，1 ，f QL3，2 ，...，f QL3，J ]，f QL3，j Represents the jth neuron output of the Q-way of the third layer of the hidden layer, J =1,2.., J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m，k，j The coefficient values of the jth neuron in the third layer of the hidden layer and the kth neuron in the second layer of the hidden layer are connected, and the initial values are all 1.
Step (8) calculating the output of neurons in the output layer
m represents the current channel I or Q, w m，k The coefficient values of the kth neuron connected with the third layer of the hidden layer are output for m paths in the output layer, the initial values are all 1, and the output f of the neural network at the current moment is obtained I And f Q 。
Step (9) outputting f I And f Q Form a complex signal f I +f Q i is sent to a decision-maker,
representing an imaginary number. The label signal s at the current time t is compared 1，t As the current output f I +f Q i, calculating error value e = s 1，t -(f I +f Q i)。
Step (10) iterative control;
in the iteration control unit, according to the current error value e and the current neural network coefficient w m，k，j (t) calculating the coefficient w of the neural network at the next moment m，k，j (t+1)＝w m，k，j (t)+Δef mk，j M denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk，j Denotes w m，k，j The output of the connected neuron.
Step (11) repeat step (3) > E(10) Until the simulation iterates all the data generated in the step (2), obtaining a group of neural network coefficients w m，k，j And storing the data in a register inside the chip.
The online operation part specifically comprises the following steps:
and (a) receiving a target signal subjected to co-channel interference by the antenna, and receiving an IQ signal through frequency conversion operation, analog-to-digital conversion and demodulation operation.
Respectively sending the IQ signals into a shift register with the time delay length of K, extracting data on the shift register and sending the data into an input layer neural network, calculating the output F (y) = y + tanh (y) of each neuron in 2K neurons of an input layer, wherein a random variable y of the input IQ signals corresponds to the data I of K shift registers of an I path y ＝[y I，1 ，y I，2 ，...，y I，K ]And Q paths of data Q of K shift registers y ＝[y Q，1 ，y Q，2 ，...，y Q，K ]；
I-way output of input layer neural network
Q path output
f IL0，k Is the k-th neuron output of the I channel, f QL0，k For the Q K-th neuron output, K =1,2.
Step (c) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the neural unit output F of the I path IL1 Neural element output F of sum Q QL1 ；F IL1 ＝[f IL1，1 ，f IL1，2 ，...，f IL1，J ]，f IL1，j Represents the jth neuron output of the first layer I path of the hidden layer, F QL1 ＝[f QL1，1 ，f QL1，2 ，...，f QL1，J ]，f QL1，j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2,. The J is the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the kth neuron of the input layer for the jth neuron in the first layer of the hidden layer, and setting the initial value as the neural network coefficient w determined in the step (11) m，k，j 。
Step (d) calculating the output of the second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the second layer of neurons is equal to the number J of the first layer of neurons, the output of the second layer of neurons comprises the output F of the neural unit of the I path IL2 And neural element output of the Q-way F QL2 ；F IL2 ＝[f IL2，1 ，f IL2，2 ，...，f IL2，J ]，f IL2，j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2，1 ，f QL2，2 ，...，f QL2，J ]，f QL2，j Represents the J-th neuron output of the Q-way of the second layer of the hidden layer, J =1,2., J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the kth neuron of the first hidden layer to the jth neuron of the second hidden layer, and setting the initial value as the neural network coefficient w determined in the step (11) m，k，j 。
Step (e) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the third layer of neurons is equal to the number J of the first layer of neurons, the output of the third layer of neurons comprises the output F of the neural unit of the path I IL3 And neural element output of the Q-way F QL3 ；F IL3 ＝[f IL3，1 ，f IL3，2 ，...，f IL3，J ]，f IL3，j Represents the jth neuron output of the third layer I of the hidden layer, F QL3 ＝[f QL3，1 ，f QL3，2 ，...，f QL3，J ]，f QL3，j Represents the jth neuron output of the Q-way of the third layer of the hidden layer, J =1,2.., J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the kth neuron of the third layer of the hidden layer with the coefficient value of the kth neuron of the second layer of the hidden layer for the jth neuron of the third layer of the hidden layer, wherein the initial value is the neural network coefficient w determined in the step (11) m，k，j 。
Step (f) computing output of neurons in the output layer
m represents the current channel I or Q, w m，k Outputting the coefficient value of the k-th neuron of the third layer of the continuous hidden layer for m paths in the output layer, wherein the initial value is the neural network coefficient w determined in the step (11) m，k，j To obtain the output f of the neural network at the current moment I And f Q 。
Step (g) outputting f I And f Q Form a complex signal f I +f Q i is sent into a decision device, and f is judged according to the standard constellation point position of the current constellation mode I +f Q i corresponds to the nearest constellation point s std The constellation mode is one of QPSK, 8PSK, 16APSK, 32APSK, 64APSK, 128APSK and 256APSK, and the constellation point s is calculated std And a complex signal f I +f Q i error e' = s std -(f I +f Q i)。
Step (h) iterative control;
in the iteration control unit, according to the current error value e' and the current neural network coefficient wm，k，j (t) calculating the coefficients of the neural network at the next moment
m denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk，j Denotes w m，k，j The output of the connected neuron.
Step (i) performs steps (b) to (h) for each of the continuously input IQ signals, the output f of the neural network at each time I And f Q Namely the output after the same frequency interference is suppressed.
The method of the invention pre-trains a group of neural network coefficients capable of resisting co-channel interference, and pre-sets the coefficients into the chip, and when the chip normally works, the pre-trained coefficients are used for resisting the influence of co-channel interference, and meanwhile, the method can also be self-adaptively adapted to the current channel state in real time. When the channel changes, the change of the upper channel can be tracked according to the constellation characteristics. Different from the traditional method for resisting same frequency interference, such as reconstruction offset, channel interference avoidance or special coding and the like, the method can tile the same frequency interference noise into the whole frequency energy without modifying the originating information standard, thereby effectively reducing the occurrence of burst error codes.
Drawings
FIG. 1 is a channel interference model of the present invention;
FIG. 2 is a schematic diagram of the structure of a single neuron according to the present invention;
FIG. 3 is a flow chart of the method of the present invention;
FIG. 4 is a diagram of a neural network architecture in the method of the present invention;
FIG. 5 is a schematic view of a portion of the on-line operation of the process of the present invention;
fig. 6 is a schematic diagram of a demodulation link in the method of the present invention.
Detailed Description
It is to be understood that the above examples are illustrative of the present invention and are not to be construed as limiting the invention, and any invention which does not depart from the spirit and scope of the invention is deemed to be within the scope and spirit of the invention.
As shown in FIG. 3, the method for suppressing co-channel interference of a digital satellite based on a neural network comprises a neural network pre-training part and an online operation part.
The neural network pre-training part specifically comprises the following steps:
step (1) generating a standard signal with length of L according to the digital satellite television broadcast signal, wherein the standard signal is QPSK and 8PSK16APSK, 32APSK, 64APSK, 128APSK, 256 APSK; random white Gaussian noise is added at the theoretical threshold of the standard signal; taking the sum of the standard signal and random white Gaussian noise as a tag signal S 1 ＝[s 1，1 ，s 1，2 ，...，s 1，L ]，s 1，l L =1,2, which is the L-th tag signal 1，l Is a plurality of numbers.
Step (2) generating co-channel interference signal S with random length L 2 ，S 2 ＝[s 2，1 ，s 2，2 ，...，s 2，L ]，s 2，l For the ith co-channel interference signal, L =1,2 2，l Is a plurality of numbers.
Will S 2 Is added to S 1 To obtain a received analog signal S 3 ＝S 1 +S 2 ，S 3 ＝[s 1，1 +s 2，1 ，s 1，2 +s 2，2 ，..，s 1，L +s 2，L ]。
Step (3) adding S 3 The real part and the imaginary part of the signal are respectively sent into two groups of independent shift registers serving as information extraction layers; at t 1 Time extraction S 3 1 to Kth data at t 2 Time of day extraction S 3 Sequentially operating the data from the 2 nd to the K +1 th according to the time sequence, and extracting the data at the L-K moments, wherein K & lt L; the data extracted at each time comprises real part data I x ＝[x I，1 ，x I，2 ，...，x I，K ]And imaginary data Q x ＝[x Q，1 ，x Q，2 ，...，x Q，K ]；
The above operation is defined as a delay shift operation of input data. In the next step, the input data is delay shifted once for each signal processing calculation.
Step (4) as shown in fig. 4, extracting data on the shift register and sending the data to the input layer neural network, calculating the output F (x) = x + tanh (x) of each neuron in 2K neurons of the input layer, where tanh (·) is a hyperbolic tangent function, and the input variable x corresponds to the data I extracted at each time x ＝[x I，1 ，x I，2 ，...，x I，K ]And Q x ＝[x Q，1 ，x Q，2 ，...，x Q，K ]；
The output of the input layer neural network comprises an I-way output F IL0 And Q path output F QL0 ：
Step (5) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the output F of the neural unit of the I path IL1 And neural element output of the Q-way F QL1 ；F IL1 ＝[f IL1，1 ，f IL1，2 ，...，f IL1，J ]，f IL1，j Represents the jth neuron output of the first layer I of the hidden layer, F QL1 ＝[f QL1，1 ，f QL1，2 ，...，f QL1，J ]，f QL1，j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2,. The J is the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m，k，j The initial values are all 1 for the coefficient values of the j-th neuron in the first layer of the hidden layer connected to the k-th neuron of the input layer.
Step (6) calculating the output of the second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two networks, and the number of the second layer of neurons and the number of the first layer of neurons areThe number J is equal, and the output comprises the neural unit output F of the way I IL2 And neural element output of the Q-way F QL2 ；F IL2 ＝[f IL2，1 ，f IL2，2 ，...，f IL2，J ]，f IL2，j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2，1 ，f QL2，2 ，...，f QL2，J ]，f QL2，j Represents the jth neuron output of the Q-way of the second layer of the hidden layer, J =1,2.., J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m，k，j The coefficient values of the jth neuron in the second layer of the hidden layer and the kth neuron in the first layer of the hidden layer are connected, and the initial values are all 1.
Step (7) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the third layer of neurons is equal to the number J of the first layer of neurons, the output of the third layer of neurons comprises the output F of the neural unit of the path I IL3 And neural element output of the Q-way F QL3 ；F IL3 ＝[f IL3，1 ，f IL3，2 ，...，f IL3，J ]，f IL3，j Represents the jth neuron output of the third layer I of the hidden layer, F QL3 ＝[f QL3，1 ，f QL3，2 ，...，f QL3，J ]，f QL3，j Represents the jth neuron output of the Q-way of the third layer of the hidden layer, J =1,2.., J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m，k，j The coefficient values of the jth neuron in the third layer of the hidden layer and the kth neuron in the second layer of the hidden layer are connected, and the initial values are all 1.
Step (8) calculating the output of neurons in the output layer
m represents the current channel I or Q, w m，k The coefficient value of the kth neuron which is connected with the third layer of the hidden layer and is output by m paths in the output layer is 1 in all initial values, namely the output f of the neural network at the current moment is obtained I And f Q 。
Step (9) outputting f I And f Q Form a complex signal f I +f Q i is sent to a decision-maker,
representing an imaginary number. The label signal s at the current time t is compared 1，t As the current output f I +f Q i, calculating error value e = s 1，t -(f I +f Q i)。
Step (10) iterative control;
in the iteration control unit, according to the current error value e and the current neural network coefficient w m，k，j (t) calculating the coefficient w of the neural network at the next moment m，k，j (t+1)＝w m，k，j (t)+Δef mk，j M denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk，j Denotes w m，k，j The output of the connected neuron.
Step (11) repeating steps (3) - (10) until the simulation iteration is finished, obtaining a group of neural network coefficients w m，k，j And storing in a register inside the chip.
As shown in fig. 5, the online operation part specifically includes:
the antenna in step (a) receives the target signal subjected to co-channel interference, and receives the IQ signal through frequency conversion, analog-to-digital conversion, and demodulation operations, as shown in fig. 6.
Respectively sending the IQ signals into a shift register with the time delay length of K, extracting data on the shift register and sending the data into an input layer neural network, calculating the output F (y) = y + tanh (y) of each neuron in 2K neurons of an input layer, wherein a random variable y of the input IQ signals corresponds to the data I of K shift registers of an I path y ＝[y I，1 ，y I，2 ，...，y I，K ]And Q paths of data Q of K shift registers y ＝[y Q，1 ，y Q，2 ，...，y Q，K ]；
I-way output of input layer neural network
Q path output
f IL0，k Is the k-th neuron output of the I channel, f QL0，k For the Q K-th neuron output, K =1,2.
Step (c) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the neural unit output F of the I path IL1 Neural element output F of sum Q QL1 ；F IL1 ＝[f IL1，1 ，f IL1，2 ，...，f IL1，J ]，f IL1，j Represents the jth neuron output of the first layer I path of the hidden layer, F QL1 ＝[f QL1，1 ，f QL1，2 ，...，f QL1，J ]，f QL1，j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2,. The J is the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the jth neuron in the first layer of the hidden layer with the coefficient value of the kth neuron of the input layer, wherein the initial value is the neural network coefficient w determined in the step (11) m，k，j 。
Step (d) calculating the output of the second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two-way network, the number of the second layer of neurons is equal to the number J of the first layer of neurons, the output of the second layer of neurons comprises the output F of the neural unit of the I way IL2 And neural element output of the Q-way F QL2 ；F IN2 ＝[f IL2，1 ，f IL2，2 ，...，f IL2，J ]，f IL2，j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2，1 ，f QL2，2 ，...，f QL2，J ]，f QL2，j Represents the jth neuron output of the Q-way of the second layer of the hidden layer, J =1,2.., J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the kth neuron of the first hidden layer to the jth neuron of the second hidden layer, and setting the initial value as the neural network coefficient w determined in the step (11) m，k，j 。
Step (e) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the third layer of neurons is equal to the number J of the first layer of neurons, the output of the third layer of neurons comprises the output F of the neural unit of the path I IL3 Neural element output F of sum Q QL3 ；F IL3 ＝[f IL3，1 ，f IL3，2 ，...，f IL3，J ]，f IL3，j Represents the jth neuron output of the I path of the third layer of the hidden layer, F QL3 ＝[f QL3，1 ，f QL3，2 ，...，f QL3，J ]，f QL3，j Represents the jth neuron output of the Q-way of the third layer of the hidden layer, J =1,2.., J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m，k，j Connecting the coefficient value of the kth neuron of the third layer of the hidden layer with the coefficient value of the kth neuron of the second layer of the hidden layer for the jth neuron of the third layer of the hidden layer, wherein the initial value is the neural network coefficient w determined in the step (11) m，k，j 。
Step (f) computing output of neurons in the output layer
m represents the current channel I or Q, w m，k For m paths of coefficient values of k-th neuron connected with the third layer of the hidden layer in the output layer, the initial value is the neural network coefficient w determined in the step (11) m，k，j To obtain the output f of the neural network at the current moment I And f Q 。
Step (g) outputting f I And f Q Form a complex signal f I +f Q i is sent into a decision device, and f is judged according to the standard constellation point position of the current constellation mode I +f Q i corresponds to the nearest constellation point s std The constellation mode is one of QPSK, 8PSK, 16APSK, 32APSK, 64APSK, 128APSK and 256APSK, and the constellation point s is calculated std And a complex signal f I +f Q i error e' = s std -(f I +f Q i)。
Step (h) iterative control;
in the iteration control unit, according to the current error value e' and the current neural network coefficient w m，k， j (t) calculating the coefficient of the neural network at the next moment
m denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk，j Denotes w m，k，j The output of the connected neuron.
Step (i) performs steps (b) to (h) for each of the continuously input IQ signals, the output f of the neural network at each time I And f Q Namely the output after the same frequency interference is restrained.
Claims (2)
1. A method for inhibiting co-channel interference by a digital satellite based on a neural network comprises a neural network pre-training part and an online operation part; the method is characterized in that:
the neural network pre-training part specifically comprises the following steps:
generating a standard signal with the length of L according to a digital satellite television broadcast signal; random white Gaussian noise is added at the theoretical threshold of the standard signal; taking the sum of standard signal and random white Gaussian noise as a label signalS 1 ＝[s 1,1 ,s 1,2 ,...,s 1,L ]，s 1,l For the ith tag signal, L =1,2, …, L, s 1,l Is a plurality;
step (2) generating co-channel interference signal S with random length L 2 ，S 2 ＝[s 2,1 ,s 2,2 ,...,s 2,L ]，s 2,l For the L co-channel interference signal, L =1,2, …, L, s 2,l Is a plurality of;
will S 2 Is added to S 1 To obtain a received analog signal S 3 ＝S 1 +S 2 ，S 3 ＝[s 1,1 +s 2,1 ,s 1,2 +s 2,2 ,..,s 1,L +s 2,L ]；
Step (3) adding S 3 The real part and the imaginary part of the signal are respectively sent into two groups of independent shift registers serving as information extraction layers; at t 1 Time of day extraction S 3 1 to Kth data at t 2 Time of day extraction S 3 Sequentially operating the data from the 2 nd to the K +1 th according to the time sequence, and extracting the data at the L-K moments, wherein K & lt L; the data extracted at each time comprises real part data I x ＝[x I,1 ,x I,2 ,…,x I,K ]And imaginary data Q x ＝[x Q,1 ,x Q,2 ,…,x Q,K ]；
Extracting data on the shift register and sending the data into an input layer neural network, calculating the output F (x) = x + tanh (x) of each neuron in 2K neurons of an input layer, wherein the tanh (·) is a hyperbolic tangent function, and an input variable x corresponds to the data I extracted at each moment x ＝[x I,1 ,x I,2 ,…,x I,K ]And Q x ＝[x Q,1 ,x Q,2 ,…,x Q,K ]；
The output of the input layer neural network comprises an I-way output F IL0 And Q path output F QL0 ：
step (5) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the output F of the neural unit of the I path IL1 And neural element output of the Q-way F QL1 ；F IL1 ＝[f IL1,1 ,f IL1,2 ,...,f IL1,J ]，f IL1,j Represents the jth neuron output of the first layer I path of the hidden layer, F QL1 ＝[f QL1,1 ,f QL1,2 ,...,f QL1,J ]，f QL1,j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2, …, J and J are the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient value of the kth neuron of the input layer for the jth neuron in the first layer of the hidden layer, wherein the initial values are all 1;
and (6) calculating the output of a second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the second layer of neurons is equal to the number J of the first layer of neurons, and the output of the second layer of neurons comprises the output F of the neural unit of the path I IL2 And neural element output of the Q-way F QL2 ；F IL2 ＝[f IL2,1 ,f IL2,2 ,...,f IL2,J ]，f IL2,j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2,1 ,f QL2,2 ,...,f QL2,J ]，f QL2,j Represents the jth neuron output of the Q-way of the second layer of the hidden layer, J =1,2, …, J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient values of the kth neuron of the first layer of the hidden layer to the jth neuron of the second layer of the hidden layer, wherein the initial values are all 1;
step (7) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two networks, the number of the third layer of neurons is equal to the number J of the first layer of neurons, and the output of the third layer of neurons comprises the output F of the neural unit of the I path IL3 And neural element output of the Q-way F QL3 ；F IL3 ＝[f IL3,1 ,f IL3,2 ,...,f IL3,J ]，f IL3,j Represents the jth neuron output of the third layer I of the hidden layer, F QL3 ＝[f QL3,1 ,f QL3,2 ,...,f QL3,J ]，f QL3,j Represents the jth neuron output of the third layer Q path of the hidden layer, J =1,2, …, J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient value of the kth neuron of the second layer of the hidden layer with the jth neuron of the third layer of the hidden layer, wherein the initial values are all 1;
step (8) computing output of neurons in the output layer
m represents the current channel I or Q, w m,k The coefficient value of the kth neuron which is connected with the third layer of the hidden layer and is output by m paths in the output layer is 1 in all initial values, namely the output f of the neural network at the current moment is obtained I And f Q ；
Step (9) outputting f I And f Q Form a complex signal f I +f Q i is sent to a decision-maker,
representing an imaginary number(ii) a The label signal s at the current time t is compared 1,t As the current output f I +f Q i, calculating error value e = s 1,t -(f I +f Q i)；
Step (10) iterative control;
in the iteration control unit, according to the current error value e and the current neural network coefficient w m,k,j (t) calculating the coefficient w of the neural network at the next moment m,k,j (t+1)＝w m,k,j (t)+Δef mk,j M denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk,j Denotes w m,k,j An output of the connected neuron;
step (11) repeating steps (3) - (10) until the simulation iteration is finished, obtaining a group of neural network coefficients w m,k,j The data is stored in a register inside the chip;
the online operation part specifically comprises the following steps:
receiving a target signal subjected to co-channel interference by an antenna, and receiving an IQ signal through frequency conversion operation, analog-to-digital conversion and demodulation operation;
respectively sending the IQ signals into a shift register with the time delay length of K, extracting data on the shift register and sending the data into an input layer neural network, calculating the output F (y) = y + tanh (y) of each neuron in 2K neurons of an input layer, wherein a random variable y of the input IQ signals corresponds to the data I of K shift registers of an I path y ＝[y I,1 ,y I,2 ,…,y I,K ]And Q paths of data Q of K shift registers y ＝[y Q,1 ,y Q,2 ,…,y Q,K ]；
I-way output of input layer neural network
Q path output
f IL0,k Is the k-th neuron output of the I channel, f QL0,k K =1,2, …, K for the Q-way kth neuron output;
step (c) calculating the output of the first layer neuron of the hidden layer, wherein the first layer neuron of the hidden layer is divided into I, Q two-path network, and the output of the first layer neuron comprises the neural unit output F of the I path IL1 And neural element output of the Q-way F QL1 ；F IL1 ＝[f IL1,1 ,f IL1,2 ,...,f IL1,J ]，f IL1,j Represents the jth neuron output of the first layer I of the hidden layer, F QL1 ＝[f QL1,1 ,f QL1,2 ,...,f QL1,J ]，f QL1,j Representing the jth neuron output of the first layer Q path of the hidden layer, wherein J =1,2, …, J and J are the number of the neurons of the first layer of the hidden layer;
implicit layer first layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient value of the kth neuron of the input layer for the jth neuron in the first layer of the hidden layer, and setting the initial value as the neural network coefficient w determined in the step (11) m,k,j ；
Step (d) calculating the output of the second layer of neurons of the hidden layer, wherein the second layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the second layer of neurons is equal to the number J of the first layer of neurons, the output of the second layer of neurons comprises the output F of the neural unit of the I path IL2 And neural element output of the Q-way F QL2 ；F IL2 ＝[f IL2,1 ,f IL2,2 ,...,f IL2,J ]，f IL2,j Represents the jth neuron output of the second layer I of the hidden layer, F QL2 ＝[f QL2,1 ,f QL2,2 ,...,f QL2,J ]，f QL2,j Represents the jth neuron output of the Q-way of the second layer of the hidden layer, J =1,2, …, J;
implicit layer second layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient value of the kth neuron of the first hidden layer to the jth neuron of the second hidden layer, with the initial value determined in step (11)Neural network coefficient w m,k,j ；
Step (e) calculating the output of the third layer of neurons of the hidden layer, wherein the third layer of neurons of the hidden layer is divided into I, Q two-path network, the number of the third layer of neurons is equal to the number J of the first layer of neurons, the output of the third layer of neurons comprises the output F of the neural unit of the path I IL3 And neural element output of the Q-way F QL3 ；F IL3 ＝[f IL3,1 ,f IL3,2 ,...,f IL3,J ]，f IL3,j Represents the jth neuron output of the third layer I of the hidden layer, F QL3 ＝[f QL3,1 ,f QL3,2 ,...,f QL3,J ]，f QL3,j Represents the jth neuron output of the third layer Q path of the hidden layer, J =1,2, …, J;
implicit layer three layer single neuron output
m represents the current channel I or Q, w m,k,j Connecting the coefficient value of the kth neuron of the third layer of the hidden layer with the coefficient value of the kth neuron of the second layer of the hidden layer for the jth neuron of the third layer of the hidden layer, wherein the initial value is the neural network coefficient w determined in the step (11) m,k,j ；
Step (f) computing output of neurons in the output layer
m represents the current channel I or Q, w m,k For m paths of coefficient values of k-th neuron connected with the third layer of the hidden layer in the output layer, the initial value is the neural network coefficient w determined in the step (11) m,k,j To obtain the output f of the neural network at the current moment I And f Q ；
Step (g) outputting f I And f Q Form a complex signal f I +f Q i is sent into a decision device, and f is judged according to the standard constellation point position of the current constellation mode I +f Q i corresponds to the nearest constellation point s std Calculating the constellation point s std And a complex signal f I +f Q i error e' = s std -(f I +f Q i)；
Step (h) iterative control;
in the iteration control unit, according to the current error value e' and the current neural network coefficient w m,k,j (t) calculating the coefficients of the neural network at the next moment
m denotes the current channel I or Q, Δ denotes a fixed iteration step, f mk,j Denotes w m,k,j An output of the connected neuron;
(ii) performing steps (b) to (h) for each continuously input IQ signal, the output f of the neural network at each time I And f Q Namely the output after the same frequency interference is suppressed.
2. The method for suppressing co-channel interference of a digital satellite based on a neural network as claimed in claim 1, wherein: the standard signal and the constellation mode are one of QPSK, 8PSK, 16APSK, 32APSK, 64APSK, 128APSK and 256 APSK.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211253844.8A CN115580340B (en) | 2022-10-13 | 2022-10-13 | Method for suppressing same-frequency interference by digital satellite based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211253844.8A CN115580340B (en) | 2022-10-13 | 2022-10-13 | Method for suppressing same-frequency interference by digital satellite based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115580340A true CN115580340A (en) | 2023-01-06 |
| CN115580340B CN115580340B (en) | 2024-07-23 |
Family
ID=84585780
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211253844.8A Active CN115580340B (en) | 2022-10-13 | 2022-10-13 | Method for suppressing same-frequency interference by digital satellite based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115580340B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116260547A (en) * | 2023-05-11 | 2023-06-13 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | System and method for inhibiting same-frequency interference |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102571659A (en) * | 2010-12-15 | 2012-07-11 | 中兴通讯股份有限公司 | Interference noise estimation and interference inhibition method and corresponding system |
| CN110601777A (en) * | 2019-08-29 | 2019-12-20 | 浙江大学 | Method for estimating satellite-ground downlink co-channel interference under low-orbit mobile satellite constellation |
| CN111126562A (en) * | 2019-11-22 | 2020-05-08 | 上海交通大学 | Target algorithm fitting method based on neural network, terminal and application |
| WO2021113314A1 (en) * | 2019-12-04 | 2021-06-10 | Massachusetts Institute Of Technology | Methods and apparatus for analog canceler tuning using neural networks |
-
2022
- 2022-10-13 CN CN202211253844.8A patent/CN115580340B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102571659A (en) * | 2010-12-15 | 2012-07-11 | 中兴通讯股份有限公司 | Interference noise estimation and interference inhibition method and corresponding system |
| CN110601777A (en) * | 2019-08-29 | 2019-12-20 | 浙江大学 | Method for estimating satellite-ground downlink co-channel interference under low-orbit mobile satellite constellation |
| CN111126562A (en) * | 2019-11-22 | 2020-05-08 | 上海交通大学 | Target algorithm fitting method based on neural network, terminal and application |
| WO2021113314A1 (en) * | 2019-12-04 | 2021-06-10 | Massachusetts Institute Of Technology | Methods and apparatus for analog canceler tuning using neural networks |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116260547A (en) * | 2023-05-11 | 2023-06-13 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | System and method for inhibiting same-frequency interference |
| CN116260547B (en) * | 2023-05-11 | 2023-09-22 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | System and method for inhibiting same-frequency interference |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115580340B (en) | 2024-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hamazumi et al. | A study of a loop interference canceller for the relay stations in an SFN for digital terrestrial broadcasting | |
| DE69609525T2 (en) | METHOD AND ARRANGEMENT FOR REDUCING NEIGHBOR INTERFERENCE USING DIVERSITY SIGNALS OF AN ANTENNA FIELD | |
| CN101558570B (en) | Improving receiver performance in a communication network | |
| DE69329893T2 (en) | Method and device for compensating for multipath fading and single-wave interference in a radio signal | |
| US8238820B2 (en) | On-channel repeater and on-channel repeating method | |
| US7187726B2 (en) | Method for demodulating a digital signal based on a generated error signal, and an associated receiver | |
| US20070191071A1 (en) | Method for isofrequency transmission of a digital signal with echo suppression and corresponding retransmission device | |
| AU2009288957A1 (en) | Channel estimation and equalization for hard-limited signals | |
| DE60131243T2 (en) | Adaptive equalization method and adaptive equalizer | |
| CN115580340A (en) | Method for inhibiting same frequency interference of digital satellite based on neural network | |
| CN113325375B (en) | Self-adaptive cancellation method based on deep neural network | |
| CN104410593B (en) | Numerical chracter nonlinearity erron amendment equalization methods based on decision-feedback model | |
| Muranov et al. | On deep learning assisted self-interference estimation in a full-duplex relay link | |
| US6985523B2 (en) | Method and system for adaptive equalization for receivers in a wide-band satellite communications system | |
| WO2020209093A1 (en) | Wireless communication system, wireless communication method, transmitting station device, and receiving station device | |
| CN105429924B (en) | A kind of Mixed cascading OFDM channel estimation method applied to satellite channel | |
| Bilbao et al. | Ai-based inter-tower communication networks: Challenges and benefits | |
| CN116132236B (en) | Single-symbol self-adaptive frequency offset estimation and compensation method applied to 5G NR system | |
| CN114697183A (en) | Channel synchronization method based on deep learning | |
| US20070217554A1 (en) | Noise Canceling in Equalized Signals | |
| Wei et al. | Single-channel demodulation algorithm for non-cooperative PCMA signals based on neural network | |
| Laot et al. | Experimental results on adaptive MMSE turbo equalization in shallow underwater acoustic communication | |
| Bilbao et al. | Improving ITCN Channel Estimation using AI-based Techniques | |
| CN116938656A (en) | FM demodulation method based on deep learning | |
| CN114884783B (en) | Method for estimating power line system channel by utilizing neural network |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Country or region after: China Address after: 310012 5-6 / F, block a, East Software Park Innovation Building, 90 Wensan Road, Hangzhou City, Zhejiang Province Applicant after: Hangzhou Guoxin Microelectronics Co.,Ltd. Address before: 310012 5-6 / F, block a, East Software Park Innovation Building, 90 Wensan Road, Hangzhou City, Zhejiang Province Applicant before: HANGZHOU NATIONALCHIP SCIENCE & TECHNOLOGY Co.,Ltd. Country or region before: China |
|
| GR01 | Patent grant |