CN113115247A - Signal processing method, device and equipment - Google Patents
Signal processing method, device and equipment Download PDFInfo
- Publication number
- CN113115247A CN113115247A CN202110283973.0A CN202110283973A CN113115247A CN 113115247 A CN113115247 A CN 113115247A CN 202110283973 A CN202110283973 A CN 202110283973A CN 113115247 A CN113115247 A CN 113115247A
- Authority
- CN
- China
- Prior art keywords
- signal sequence
- attention
- original signal
- value information
- verified
- 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
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information
-
- 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
- G08—SIGNALLING
- G08C—TRANSMISSION SYSTEMS FOR MEASURED VALUES, CONTROL OR SIMILAR SIGNALS
- G08C17/00—Arrangements for transmitting signals characterised by the use of a wireless electrical link
- G08C17/02—Arrangements for transmitting signals characterised by the use of a wireless electrical link using a radio link
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/06—Optimizing the usage of the radio link, e.g. header compression, information sizing, discarding information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/02—Power saving arrangements
- H04W52/0209—Power saving arrangements in terminal devices
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The application relates to a signal processing method, a device and equipment, wherein the method comprises the following steps: collecting an original signal sequence in a preset period; compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence; extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence; and sending the compressed signal sequence and the attention characteristics to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value. Therefore, the power consumption of the wireless sensor node can be reduced, the working time of the sensor can be prolonged, and the accuracy of signal transmission and reconstruction can be improved.
Description
Technical Field
The present application relates to the field of medical equipment technologies, and in particular, to a signal processing method, device and equipment.
Background
As an important branch of the internet of things, body sensor networks (also called "somatosensory networks", "wearable sensor networks", etc.) have been widely used in recent years, such as physiological parameter monitoring, chronic disease management, health wrist watches, fall monitoring, etc. However, in a scene requiring real-time continuous acquisition, how to reduce the sampling and transmission power consumption of the wireless sensor node and extend the working time of the sensor node is a bottleneck problem to be broken through.
It is known that, in the framework of conventional digital signal processing based on Shannon/Nyquist sampling theorem, if the original signal is recovered from the sampled discrete signal without distortion, the sampling frequency must be more than twice its bandwidth.
However, in the field of body sensor networks, especially in the scene requiring real-time continuous acquisition (such as real-time electrocardiographic monitoring, real-time motion capture, etc.), since a large amount of data needs to be transmitted in a wireless manner, the power consumption of the wireless sensor node is always high, and the working time is greatly shortened.
Disclosure of Invention
The embodiment of the application provides a signal processing method, a signal processing device and signal processing equipment, which can reduce the power consumption of a wireless sensor node, prolong the working time of a sensor and improve the accuracy of signal transmission and reconstruction.
In one aspect, an embodiment of the present application provides a signal processing method, which is applied to an acquisition end of a body sensor network, and the method includes:
collecting an original signal sequence in a preset period;
compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence;
extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence;
and sending the compressed signal sequence and the attention characteristics to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
Optionally, the trained data compression model sequentially includes a first convolution layer, a batch normalization layer, a first activation layer, a second convolution layer, a batch normalization layer, a second activation layer, a third convolution layer, a maximum pooling layer, a batch normalization layer, and a third activation layer;
the input end of the first convolution layer is used for inputting an original signal sequence, and the output end of the third activation layer is used for outputting a compressed signal sequence.
Optionally, the size of the convolution kernel in the first convolution layer, the second convolution layer, and the third convolution layer is 1 × n, and the step size ranges from 1 to n, where n is 2m +1, and m is 0,1,2 … ….
Optionally, the trained feature extraction model includes an extreme attention module and an attention module; the attention feature includes extreme value information and attention value information;
extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence, wherein the method comprises the following steps:
carrying out extremum feature extraction on the original signal sequence through an extremum attention module to obtain extremum information corresponding to the original signal sequence;
and performing attention value feature extraction on the original signal sequence through an attention value attention module to obtain attention value information corresponding to the original signal sequence.
On the other hand, an embodiment of the present application further provides a signal processing method, including:
receiving a compressed signal sequence and attention characteristics sent by an acquisition end of a body sensor network; the compressed signal sequence is obtained by compressing an original signal sequence acquired in a preset period by an acquisition end according to a trained data compression model; the attention feature is obtained by the acquisition end extracting the features of the original signal sequence according to the trained feature extraction model;
reconstructing the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence;
and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
Optionally, the attention feature includes extreme value information and attention value information; the extreme value information and the attention value information respectively carry corresponding sampling time identifications;
checking the reconstructed signal sequence according to the attention characteristics, comprising:
determining a signal to be checked from the reconstructed signal sequence based on the sampling time identifier carried by the extreme value information, the sampling time identifier carried by the attention value information and the sampling time identifier corresponding to each reconstructed signal in the reconstructed signal sequence;
comparing the amplitude corresponding to the signal to be verified with the amplitude corresponding to the extreme value information and the amplitude corresponding to the attention value information;
and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information is in a preset range, or the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information is in a preset range, determining that the reconstructed signal sequence and the original signal sequence after verification meet a preset matching degree value.
Optionally, the method further comprises:
and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the extreme value information.
Or; and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the attention value information.
On the other hand, this application embodiment still provides a signal processing apparatus, is applied to body sensor network's collection end, includes:
the acquisition module is used for acquiring an original signal sequence in a preset period;
the first determining module is used for compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence;
the second determining module is used for extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence;
and the sending module is used for sending the compressed signal sequence and the attention characteristics to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifies the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
On the other hand, an embodiment of the present application further provides a signal processing apparatus, including:
the receiving module is used for receiving a compressed signal sequence and attention characteristics sent by an acquisition end of the body sensor network; the compressed signal sequence is obtained by compressing an original signal sequence acquired in a preset period by an acquisition end according to a trained data compression model; the attention feature is obtained by the acquisition end extracting the features of the original signal sequence according to the trained feature extraction model;
the first determining module is used for reconstructing the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence;
and the second determining module is used for verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
On the other hand, the embodiment of the present application further provides an apparatus, where the apparatus includes a processor and a memory, where at least one instruction or at least one program is stored in the memory, and the at least one instruction or the at least one program is loaded by the processor and executes the signal processing method described above.
The signal processing method, the signal processing device and the signal processing equipment have the following beneficial effects:
acquiring an original signal sequence in a preset period; compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence; extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence; and sending the compressed signal sequence and the attention characteristics to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value. Therefore, the power consumption of the wireless sensor node can be reduced, the working time of the sensor can be prolonged, and the accuracy of signal transmission and reconstruction can be improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic diagram of an application scenario provided in an embodiment of the present application;
fig. 2 is a schematic flowchart of a signal processing method according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of a data compression model provided in an embodiment of the present application;
FIG. 4 is a diagram illustrating convolution kernel size and signal according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a one-dimensional raw signal sequence attention heat map provided by an embodiment of the present application;
FIG. 6 is a schematic diagram of an attention thermal map encoding process provided by an embodiment of the present application;
fig. 7 is a schematic structural diagram of a signal processing apparatus according to an embodiment of the present application;
fig. 8 is a schematic structural diagram of a signal processing apparatus according to an embodiment of the present application;
fig. 9 is a block diagram of a hardware structure of a server according to a signal processing method provided in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, 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 application.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In the field of the existing body sensor network, under the scene of continuously acquiring motion signals in real time, a large amount of data needs to be transmitted in a wireless mode, so that the power consumption of wireless sensor nodes is always high.
Referring to fig. 1, fig. 1 is a schematic diagram of an application scenario provided in an embodiment of the present application, including an acquisition end 101 and a receiving end 102; the acquisition end 101 and the receiving end 102 are in the same body sensing network, and the acquisition end 101 comprises a sensor 1011, a preprocessing module 1012 and a sending module 1013; wherein, the sensor 1011 is used for collecting the original signal sequence of the human body; the original signal sequence acquired by the sensor 1011 is compressed and feature extracted by the preprocessing module 1012, and then is sent to the receiving end 102 through the sending module 1013.
Specifically, firstly, an original signal sequence in a preset period is acquired through a sensor 1011 of an acquisition end 101; secondly, the original signal sequence is compressed according to the trained data compression model in the preprocessing module 1012 to obtain a compressed signal sequence; extracting the features of the original signal sequence according to the trained feature extraction model in the preprocessing module 1012 to obtain the attention features corresponding to the original signal sequence; secondly, a compressed signal sequence and attention characteristics are sent to the receiving end 102 of the body sensor network through the sending module 1013; after receiving the compressed signal sequence and the attention feature, the receiving end 102 reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifies the reconstructed signal sequence according to the attention feature so that the verified reconstructed signal sequence and the original signal sequence satisfy a preset matching degree value.
Optionally, the body sensor network includes 3 or more than 3 acquisition terminals 101.
Optionally, the sensor 1011 may be a sensor for collecting human body motion information, such as an acceleration sensor, and when being worn on the upper arm and the forearm of the human body, the sensor may collect the limb motion information of the human body; the sensor 1011 may also be a sensor for collecting physiological signals (such as ecg, eeg, blood oxygen signals, etc.) of a human body.
Optionally, the data compression model, the feature extraction model in the preprocessing module 1012, and the data reconstruction model in the receiving end 102 all use a Convolutional Neural Network (CNN) as a basic structure. Thus, different from the conventional data compression method, the present application uses a combination of multiple filters in hardware to achieve a data compression function, and performs data compression (encoding) at the acquisition end 101 through a CNN-based data compression model, and at the same time, since the designed data compression model is determined, the reception end 102 performs data reconstruction (decoding) through a constructed corresponding data reconstruction model.
While specific embodiments of a signal processing method according to the present application are described below, fig. 2 is a schematic flow chart of a signal processing method according to the embodiments of the present application, and the present specification provides the method operation steps according to the embodiments or the flow chart, but more or less operation steps may be included based on conventional or non-inventive labor. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of orders and does not represent the only order of execution. In practice, the system or server product may be implemented in a sequential or parallel manner (e.g., parallel processor or multi-threaded environment) according to the embodiments or methods shown in the figures. Specifically, as shown in fig. 2, the method may include:
s201: and the acquisition end acquires an original signal sequence in a preset period.
In the embodiment of the application, the sensor serves as a collection node, and collects an original signal sequence of a human body according to a preset period, wherein the original signal sequence may refer to a physiological signal sequence of the human body or a motion signal sequence, for example, a limb motion signal collected by being worn on an upper arm and a forearm of the human body during rehabilitation training of a patient. Compared with the prior art that most of signal processing methods related to reducing the power consumption of the wireless sensing node can only process human physiological signals, and for human motion signals, the human motion signals are low-frequency (mainly concentrated in the range of 3-20 Hz) one-dimensional observation sequences, and the motion signals between different parts have weak correlation.
S203: and the acquisition end compresses the original signal sequence according to the trained data compression model to obtain a compressed signal sequence.
In the embodiment of the application, the data compression model adopts a CNN convolutional network as a basic structure. The CNN convolution kernel is essentially a filter, and the CNN convolution kernel function is realized in hardware by designing a plurality of specific filters, and the signal compression based on the CNN network is realized in hardware by cascading the filters.
In an alternative embodiment, as shown in fig. 3, the trained data compression model sequentially includes a first convolution layer, a batch normalization layer, a first activation layer, a second convolution layer, a batch normalization layer, a second activation layer, a third convolution layer, a maximum pooling layer, a batch normalization layer, and a third activation layer; the input end of the first convolution layer is used for inputting an original signal sequence, and the output end of the third activation layer is used for outputting a compressed signal sequence.
In this optional embodiment, the size of the convolution kernel is directly related to the system signal characteristics and the sampling rate, and the size of the convolution kernel represents the window width of signal characteristic extraction; as shown in fig. 4, the size of the convolution kernel in each convolution layer can be represented by general formula 1 × n, where n is 2m +1, m is 0,1,2 … …, and the sizes of the convolution kernels may or may not be all the same between each convolution layer; the range of the convolution kernel moving step length (stride) is 1-n, signal feature extraction is missed when the range is exceeded, and the system calculation amount is overlarge when the range is too small; preferably, m +1 is used as the convolution kernel moving step size, namely, the overlap between adjacent convolution kernels is half, and the subsequent convolution kernel extracts the features from the center of the previous convolution kernel.
Considering that the human motion signal period is 3-20Hz, and the general sampling rate is less than 40Hz, therefore, the applicable CNN convolution kernels can include 1 × 3, 1 × 5 and 1 × 7, the smaller the convolution kernel is, the better the feature extraction effect is, and the more accurate the data compression is; cardiac electrical signals (typically at a sampling rate of 250Hz), and corresponding, applicable convolution kernels may comprise 1 x 5 or 1 x 7; the electromyographic signals (the main energy is concentrated in 20-150 Hz, and the common sampling rate is 300Hz), and the corresponding convolution kernels can comprise 1 × 5, 1 × 7 and the like.
In a specific embodiment, when used for acquiring a human motion signal, the convolution kernels in the first convolution layer have a size of 1 x1 and a number of 16; the convolution kernels in the second convolution layer are 1 x3 in size and 16 in number; the convolution kernels in the third convolution layer are 1 x1 in size and 32 in number; the size of the pooling core in the largest pooling layer is 1 x3, so that the nonlinear dimensionality reduction of the parameters can be reduced; in addition, dropout can also be added at the end of each layer (before or after the activation layer) to reduce model parameters.
It should be noted that the above model structure is not the only way that the present application can be implemented, and in other embodiments of the present application, the data compression model may further include modules such as a fourth convolution layer, a fifth convolution layer, or a full link layer, so as to obtain a better compression effect; in addition, the size and the scale of the convolution kernel in each convolution layer can be set according to the actually acquired original signal sequence, the convolution kernel can adopt a symmetric convolution kernel or an asymmetric convolution kernel, and different convolution kernels in the same convolution layer can also adopt different sizes so as to obtain information of different scales; for example, when aiming at some complex human physiological signals, the data compression model and the structure of a feature extraction model to be mentioned below can be designed according to actual needs, the sizes of convolution kernels in each convolution layer can be redesigned according to actual needs, the larger the convolution kernel is, the larger the receptive field is, so that the interference caused by too much local information can be reduced, and the parameter settings of other layers are synchronously optimized along with the different sizes of the convolution kernels.
Optionally, the first active layer, the second active layer, and the third active layer may each employ an exponential linear unit as an activation function, such as the following equation:
s205: and the acquisition end performs characteristic extraction on the original signal sequence according to the trained characteristic extraction model to obtain the attention characteristic corresponding to the original signal sequence.
In the embodiment of the application, considering that some characteristics are of special concern during the movement of the patient, specific characteristics of the human body movement signal (such as the highest point, the lowest point, the movement speed and the like of the limb movement) have significance for the estimation of the limb movement function, and the loss or deviation of part of information in the compression process is unacceptable; therefore, the method and the device construct a feature extraction model based on the above idea to perform feature extraction on the original signal sequence to obtain attention features corresponding to the original signal sequence, wherein the attention features include specific features corresponding to different motion signals.
In an alternative embodiment, the trained feature extraction model includes an extreme attention module and a focus value attention module; the attention feature includes extreme value information and attention value information; the extreme value information represents a local maximum value and a local minimum value in an original signal sequence; the interest value information can refer to different meanings in different application scenarios, and here, the interest value information refers to a characteristic value near a zero crossing point in an original signal sequence, and the size of the zero crossing point is set according to actual conditions.
Correspondingly, in an optional embodiment of performing feature extraction on an original signal sequence according to a trained feature extraction model to obtain an attention feature corresponding to the original signal sequence, the method includes: carrying out extremum feature extraction on the original signal sequence through an extremum attention module to obtain extremum information corresponding to the original signal sequence; and performing attention value feature extraction on the original signal sequence through an attention value attention module to obtain attention value information corresponding to the original signal sequence.
In particular, during the movement of the patient, some characteristics are of particular interest, such as the highest and lowest position of the patient's hand, the speed of the patient's movement, and the like; when the original signal sequence is obtained, the sight line of people firstly falls on the highest point and the lowest point of the original signal sequence, and if a standard line is drawn in a graph, the sight line of a user also falls on the intersection of the original signal sequence and the standard line; when the present application performs feature extraction on an original signal sequence according to a trained feature extraction model, an area containing attention features is extracted first, which is referred to as an attention heat map in the present application, as shown in fig. 5, fig. 5 is a schematic diagram of a one-dimensional original signal sequence attention heat map provided in an embodiment of the present application, where the diagram contains an original signal sequence (shown by a black line), attention features (indicated by a circle in fig. 5) subjected to feature extraction, and a reconstructed signal sequence (shown by a gray line), the original signal sequence is compressed and then generates a new data packet with the attention features, and the new data packet is transmitted to a receiving end through wireless communication, and the receiving end verifies the attention features to generate a reconstructed signal sequence; as shown in fig. 6, the feature extraction model encodes an original signal sequence [ x1, x2, x3, … …, xn ] by using a specific CNN convolution kernel corresponding to different weight values w1, w2, w3, and w4, so as to obtain an attention heat map, and then obtains attention features from the attention heat map, where the attention features are used for packet loss and misalignment correction of subsequent wireless data transmission and accuracy check of a subsequent receiving end in a data reconstruction process.
S207: and the acquisition end sends the compressed signal sequence and the attention characteristics to the receiving end.
S209: and the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence.
S211: and the receiving end verifies the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
In the embodiment of the application, an acquisition end sends a compressed signal sequence and attention characteristics to a receiving end of a body sensor network, and the receiving end reconstructs the compressed signal sequence according to a trained data reconstruction model to obtain a reconstructed signal sequence; in essence, the data reconstruction process is the inverse operation of the data compression process, in the application, the acquisition end adopts a plurality of filters to realize the CNN convolution kernel function so as to realize data compression, and when the data of the receiving end is resolved, only the deconvolution operation needs to be carried out on a compressed signal sequence; the CNN convolution kernel of the lower computer filter designed in the data compression process is determined to be unchanged, and correspondingly, when the upper computer performs compression calculation, the corresponding deconvolution kernel can be directly designed according to the CNN convolution kernel of the lower computer. Finally, considering signal transmission errors and resolving errors, the receiving end also verifies the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value, and therefore the accuracy of signal resolving is guaranteed.
In an alternative embodiment, the attention characteristics include extremum information and attention value information; the extreme value information and the attention value information respectively carry corresponding sampling time identifications; the step of checking the reconstructed signal sequence according to the attention characteristics may include:
firstly, determining a signal to be checked from a reconstructed signal sequence based on a sampling time identifier carried by extreme value information, a sampling time identifier carried by attention value information and a sampling time identifier corresponding to each reconstructed signal in the reconstructed signal sequence; specifically, a sampling time identifier which is the same as the sampling time identifier carried by the extremum information is found from the reconstructed signal sequence, and the reconstructed signal corresponding to the same sampling time identifier is used as a signal to be checked;
secondly, comparing the amplitude corresponding to the signal to be verified with the amplitude corresponding to the extreme value information and the amplitude corresponding to the attention value information; here, the form of the amplitude is determined according to the type of the actual sensor, and may be compared with the detected angle data of the patient raising the hand, or may be compared with the detected movement speed of the patient;
if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information is in a preset range, or the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information is in a preset range, determining that the reconstructed signal sequence and the original signal sequence after verification meet a preset matching degree value; the preset range and the preset matching degree value may set the mapping relationship in advance, for example, the preset matching degree value may be 1 when the preset range is strict, that is, the fault-tolerant range is small; under the condition that the preset range is relatively loose, namely the fault-tolerant range is relatively large, the preset matching degree value can be 0.8; or, in order to simplify the calculation, fixed values are respectively set for the preset range and the preset matching degree value;
if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the extreme value information; or; if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the attention value information; specifically, when a reconstructed signal has a large error, the attention characteristics sent by the acquisition end can be directly used to replace the reconstructed signal which fails to be checked, so that the problems of packet loss and dislocation in the signal transmission process can be solved, and the accuracy of signal reconstruction can be improved.
The embodiment of the present application further provides a signal processing apparatus, which is applied to an acquisition end of a body sensor network, and fig. 7 is a schematic structural diagram of the signal processing apparatus provided in the embodiment of the present application, and as shown in fig. 7, the apparatus includes:
an acquisition module 701, configured to acquire an original signal sequence in a preset period;
a first determining module 702, configured to compress an original signal sequence according to a trained data compression model to obtain a compressed signal sequence;
a second determining module 703, configured to perform feature extraction on the original signal sequence according to the trained feature extraction model to obtain an attention feature corresponding to the original signal sequence;
the sending module 704 is configured to send the compressed signal sequence and the attention feature to a receiving end of the body sensor network, so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifies the reconstructed signal sequence according to the attention feature, so that the verified reconstructed signal sequence and the original signal sequence satisfy a preset matching degree value.
In an optional embodiment, the trained data compression model sequentially includes a first convolution layer, a batch normalization layer, a first activation layer, a second convolution layer, a batch normalization layer, a second activation layer, a third convolution layer, a maximum pooling layer, a batch normalization layer, and a third activation layer; the input end of the first convolution layer is used for inputting an original signal sequence, and the output end of the third activation layer is used for outputting a compressed signal sequence.
In an optional embodiment, the size of the convolution kernel in the first convolution layer, the second convolution layer, and the third convolution layer is 1 × n, and the step size ranges from 1 to n, where n is 2m +1, and m is 0,1,2 … ….
In an alternative embodiment, the trained feature extraction model includes an extreme attention module and a focus value attention module; the attention feature includes extreme value information and attention value information; the second determining module 703 is specifically configured to:
carrying out extremum feature extraction on the original signal sequence through an extremum attention module to obtain extremum information corresponding to the original signal sequence; and performing attention value feature extraction on the original signal sequence through an attention value attention module to obtain attention value information corresponding to the original signal sequence.
The embodiment of the present application further provides a signal processing apparatus, which is applied to a receiving end in a body sensor network, and fig. 8 is a schematic structural diagram of the signal processing apparatus provided in the embodiment of the present application, and as shown in fig. 8, the apparatus includes:
the receiving module 801 is used for receiving a compressed signal sequence and attention characteristics sent by an acquisition end of a body sensor network; the compressed signal sequence is obtained by compressing an original signal sequence acquired by the motion sensor in a preset period according to the trained data compression model by the acquisition end; the attention feature is obtained by the acquisition end extracting the features of the original signal sequence according to the trained feature extraction model;
a first determining module 802, configured to reconstruct the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence;
the second determining module 803 is configured to verify the reconstructed signal sequence according to the attention characteristics, so that the verified reconstructed signal sequence and the original signal sequence satisfy a preset matching degree value.
In an alternative embodiment, the attention feature includes extremum information and attention value information; the extreme value information and the attention value information respectively carry corresponding sampling time identifications; the second determining module 803 is specifically configured to:
determining a signal to be checked from the reconstructed signal sequence based on the sampling time identifier carried by the extreme value information, the sampling time identifier carried by the attention value information and the sampling time identifier corresponding to each reconstructed signal in the reconstructed signal sequence; comparing the amplitude corresponding to the signal to be verified with the amplitude corresponding to the extreme value information and the amplitude corresponding to the attention value information; and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information is in a preset range, or the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information is in a preset range, determining that the reconstructed signal sequence and the original signal sequence after verification meet a preset matching degree value.
In an optional implementation manner, the second determining module 803 is further specifically configured to: and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the extreme value information. Or; and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the attention value information.
The device and method embodiments in the embodiments of the present application are based on the same application concept.
The method provided by the embodiment of the application can be executed in a computer terminal, a server or a similar operation device. Taking the example of the application running on a server, fig. 9 is a block diagram of a hardware structure of the server of the signal processing method provided in the embodiment of the present application. As shown in fig. 9, the server 900 may have a relatively large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 910 (the processor 910 may include but is not limited to a Processing device such as a microprocessor NCU or a programmable logic device FPGA), a memory 930 for storing data, and one or more storage media 920 (e.g., one or more mass storage devices) for storing applications 923 or data 922. Memory 930 and storage media 920 may be, among other things, transient or persistent storage. The program stored in the storage medium 920 may include one or more modules, each of which may include a series of instruction operations in a server. Still further, the central processor 910 may be configured to communicate with the storage medium 920, and execute a series of instruction operations in the storage medium 920 on the server 900. The server 900 may also include one or more power supplies 960, one or more wired or wireless network interfaces 950, one or more input-output interfaces 940, and/or one or more operating systems 921, such as Windows, Mac OS, Unix, Linux, FreeBSD, etc.
The input/output interface 940 may be used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the server 900. In one example, the input/output Interface 940 includes a Network adapter (NIC) that can be connected to other Network devices through a base station to communicate with the internet. In one example, the input/output interface 940 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.
It will be understood by those skilled in the art that the structure shown in fig. 9 is only an illustration and is not intended to limit the structure of the electronic device. For example, server 900 may also include more or fewer components than shown in FIG. 9, or have a different configuration than shown in FIG. 9.
As can be seen from the embodiments of the signal processing method, the signal processing device, and the signal processing apparatus provided by the present application, in the present application, an original signal sequence in a preset period is acquired; compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence; extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence; and sending the compressed signal sequence and the attention characteristics to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence, and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value. Therefore, the power consumption of the wireless sensor node can be reduced, the working time of the sensor can be prolonged, and the accuracy of signal transmission and reconstruction can be improved.
It should be noted that: the sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. A signal processing method is applied to an acquisition end of a body sensor network, and comprises the following steps:
collecting an original signal sequence in a preset period;
compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence;
extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence;
and sending the compressed signal sequence and the attention feature to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to a trained data reconstruction model to obtain a reconstructed signal sequence, and verifying the reconstructed signal sequence according to the attention feature so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
2. The method of claim 1, wherein the trained data compression model comprises, in order, a first convolution layer, a batch normalization layer, a first activation layer, a second convolution layer, a batch normalization layer, a second activation layer, a third convolution layer, a max pooling layer, a batch normalization layer, and a third activation layer;
the input end of the first convolution layer is used for inputting the original signal sequence, and the output end of the third activation layer is used for outputting the compressed signal sequence.
3. The method of claim 2, wherein the convolution kernel in the first convolution layer, the second convolution layer, and the third convolution layer has a size of 1 n and a step size ranging from 1 to n, where n is 2m +1 and m is 0,1,2 … ….
4. The method of claim 1, wherein the trained feature extraction model comprises an extreme attention module and a focus value attention module; the attention feature comprises extreme value information and attention value information;
the extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence includes:
carrying out extremum feature extraction on the original signal sequence through the extremum attention module to obtain extremum information corresponding to the original signal sequence;
and performing attention value feature extraction on the original signal sequence through the attention value attention module to obtain attention value information corresponding to the original signal sequence.
5. A signal processing method, comprising:
receiving a compressed signal sequence and attention characteristics sent by an acquisition end of a body sensor network; the compressed signal sequence is obtained by compressing an original signal sequence acquired in a preset period by the acquisition end according to a trained data compression model; the attention feature is obtained by the acquisition end performing feature extraction on the original signal sequence according to a trained feature extraction model;
reconstructing the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence;
and verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
6. The method of claim 5, wherein the attention characteristics include extremum information and attention value information; the extreme value information and the attention value information respectively carry corresponding sampling time identifications;
the verifying the reconstructed signal sequence according to the attention feature includes:
determining a signal to be checked from the reconstruction signal sequence based on the sampling time identifier carried by the extreme value information, the sampling time identifier carried by the attention value information and the sampling time identifier corresponding to each reconstruction signal in the reconstruction signal sequence;
comparing the amplitude corresponding to the signal to be verified with the amplitude corresponding to the extreme value information and the amplitude corresponding to the attention value information;
and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information is in a preset range, or the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information is in a preset range, determining that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
7. The method of claim 6, further comprising:
and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the extreme value information exceeds a preset range, updating the amplitude of the signal to be verified to be the amplitude corresponding to the extreme value information.
Or; and if the error between the amplitude corresponding to the signal to be verified and the amplitude corresponding to the attention value information exceeds a preset range, updating the amplitude of the signal to be verified to the amplitude corresponding to the attention value information.
8. A signal processing device is characterized in that the signal processing device is applied to an acquisition end of a body sensor network and comprises:
the acquisition module is used for acquiring an original signal sequence in a preset period;
the first determining module is used for compressing the original signal sequence according to the trained data compression model to obtain a compressed signal sequence;
the second determining module is used for extracting the features of the original signal sequence according to the trained feature extraction model to obtain the attention features corresponding to the original signal sequence;
the sending module is used for sending the compressed signal sequence and the attention feature to a receiving end of the body sensor network so that the receiving end reconstructs the compressed signal sequence according to a trained data reconstruction model to obtain a reconstructed signal sequence, and verifies the reconstructed signal sequence according to the attention feature so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
9. A signal processing apparatus, characterized by comprising:
the receiving module is used for receiving a compressed signal sequence and attention characteristics sent by an acquisition end of the body sensor network; the compressed signal sequence is obtained by compressing an original signal sequence acquired in a preset period by the acquisition end according to a trained data compression model; the attention feature is obtained by the acquisition end performing feature extraction on the original signal sequence according to a trained feature extraction model;
the first determining module is used for reconstructing the compressed signal sequence according to the trained data reconstruction model to obtain a reconstructed signal sequence;
and the second determining module is used for verifying the reconstructed signal sequence according to the attention characteristics so that the verified reconstructed signal sequence and the original signal sequence meet a preset matching degree value.
10. An apparatus comprising a processor and a memory, wherein at least one instruction or at least one program is stored in the memory, and wherein the at least one instruction or the at least one program is loaded by the processor and executes the signal processing method according to any one of claims 1 to 4 or 5 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110283973.0A CN113115247B (en) | 2021-03-17 | 2021-03-17 | Signal processing method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110283973.0A CN113115247B (en) | 2021-03-17 | 2021-03-17 | Signal processing method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113115247A true CN113115247A (en) | 2021-07-13 |
| CN113115247B CN113115247B (en) | 2022-07-12 |
Family
ID=76711701
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110283973.0A Active CN113115247B (en) | 2021-03-17 | 2021-03-17 | Signal processing method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113115247B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115119052A (en) * | 2022-04-29 | 2022-09-27 | 河海大学 | Image data compression method and system based on attention mechanism and spatial redundancy |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104933846A (en) * | 2015-06-04 | 2015-09-23 | 中国科学院苏州生物医学工程技术研究所 | Body sensor network system based on compressed sensing |
| CN107634937A (en) * | 2017-08-29 | 2018-01-26 | 中国地质大学（武汉） | A kind of wireless sense network data compression method, equipment and its storage device |
| CN110519233A (en) * | 2019-07-31 | 2019-11-29 | 中国地质大学（武汉） | A kind of spaceborne Sensor Network data compression method based on artificial intelligence |
| CN111340238A (en) * | 2020-03-12 | 2020-06-26 | 中南大学 | Fault diagnosis method, device, equipment and storage medium of industrial system |
| CN112036467A (en) * | 2020-08-27 | 2020-12-04 | 循音智能科技(上海)有限公司 | Abnormal heart sound identification method and device based on multi-scale attention neural network |
-
2021
- 2021-03-17 CN CN202110283973.0A patent/CN113115247B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104933846A (en) * | 2015-06-04 | 2015-09-23 | 中国科学院苏州生物医学工程技术研究所 | Body sensor network system based on compressed sensing |
| CN107634937A (en) * | 2017-08-29 | 2018-01-26 | 中国地质大学（武汉） | A kind of wireless sense network data compression method, equipment and its storage device |
| CN110519233A (en) * | 2019-07-31 | 2019-11-29 | 中国地质大学（武汉） | A kind of spaceborne Sensor Network data compression method based on artificial intelligence |
| CN111340238A (en) * | 2020-03-12 | 2020-06-26 | 中南大学 | Fault diagnosis method, device, equipment and storage medium of industrial system |
| CN112036467A (en) * | 2020-08-27 | 2020-12-04 | 循音智能科技(上海)有限公司 | Abnormal heart sound identification method and device based on multi-scale attention neural network |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115119052A (en) * | 2022-04-29 | 2022-09-27 | 河海大学 | Image data compression method and system based on attention mechanism and spatial redundancy |
| CN115119052B (en) * | 2022-04-29 | 2023-10-24 | 河海大学 | Image data compression method and system based on attention mechanism and spatial redundancy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113115247B (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10575788B2 (en) | Compressive sensing of quasi-periodic signals using generative models | |
| CN110353649B (en) | Heart rate detection method | |
| US20130317377A1 (en) | Generative Model-Driven Resource-Efficient Monitoring in Body Sensor Networks | |
| Agarwal et al. | Input-feature correlated asynchronous analog to information converter for ECG monitoring | |
| Deshpande et al. | Iot based real time ecg monitoring system using cypress wiced | |
| Kew et al. | Variable threshold method for ECG R-peak detection | |
| CN111669409B (en) | Sign data monitoring system | |
| CN113115247B (en) | Signal processing method, device and equipment | |
| JP2007090000A (en) | Method for compressing electrocardiogram data, and electrocardiogram telemeter | |
| CN106788885B (en) | Electrocardiosignal compression rate self-adaptive adjustment wireless transmission system and transmission method | |
| CN110960211A (en) | Embedded-based real-time electrocardio monitoring system | |
| CN106974640A (en) | A kind of remote heart monitoring system | |
| Lal et al. | Compressed sensing approach for physiological signals: A review | |
| KR100967994B1 (en) | Method for a Real Time Compressing Transmission and Storage of Periodic Bio-signal Data, and Cardiovascular System therefor | |
| CN110537907B (en) | Electrocardiosignal compression and identification method based on singular value decomposition | |
| Sahu et al. | Diagnosis-steganography-transmission: an innovative integrated paradigm for ECG healthcare | |
| Bhowmick et al. | IoT assisted real time PPG monitoring system for health care application | |
| Hassanuzzaman et al. | End to end solution for continuous monitoring and real-time analysis of vital signs from ecg signal | |
| Sá et al. | Heart disease detection architecture for lead I off-the-person ECG monitoring devices | |
| CN113693610A (en) | Few-lead electrocardiogram data processing method and device, storage medium and computer equipment | |
| KR101526774B1 (en) | A method for transmitting an electrocardiography signal using compressed sensing and a monitoring system using thereof | |
| JP2023527742A (en) | System and method for monitoring multiple biosignals, and gateway device operable in the system | |
| Rodriguez et al. | Wireless ECG based on Bluetooth protocol: design and implementation | |
| Chang et al. | Microcontroller implementation of low-power compression for wearable biosignal transmitter | |
| Ryzhakov et al. | Electrocardiogram Signals Digital Processing in a Distributed Computing System |
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 |