Bioelectric signal separation and noise reduction method, device and electronic device

Abstract

The present invention provides a method, device and electronic device for separating and denoising bioelectric signals, and relates to the technical field of separating and denoising bioelectric signals, including: selecting a target audio separation processing model, and optimizing the target audio separation processing model based on a bioelectric signal data set to obtain a first target processing model; based on the separation processing of the current bioelectric signal, optimizing the first target processing model again to obtain a second target processing model; and separating and processing the bioelectric signal to be tested according to the second target processing model to alleviate the technical problem of insufficient data set for the bioelectric signal.

Description

Technical Field

[0001] The present invention relates to the technical field of bioelectric signal separation and noise reduction, and in particular to a bioelectric signal separation and noise reduction method, device and electronic equipment. Background Art

[0002] Bioelectric signals (electrocardiogram, electroencephalogram, electromyogram, etc.) have been widely used in clinical practice for a long time. Scientists are constantly researching new algorithms. In order to obtain higher quality signals, they can reduce the noise of bioelectric signals. At the same time, with the development of artificial intelligence, intelligent diagnosis is gradually becoming more and more mature. This technology is also naturally applied to bioelectric signals.

[0003] However, in actual application, due to the complex application environment, bioelectric signals are easily interfered with, which in turn affects the results of intelligent diagnosis. Summary of the invention

[0004] The purpose of the present invention is to provide a method, device and electronic device for separating and denoising bioelectric signals, so as to alleviate the technical problem of inaccurate diagnosis results of bioelectric signals.

[0005] In a first aspect, an embodiment of the present invention provides a method for separating and denoising a bioelectric signal, comprising:

[0006] Selecting a target audio separation processing model, and optimizing the target audio separation processing model based on the bioelectric signal data set to obtain a first target processing model;

[0007] Based on the separation processing of the current bioelectric signal, the first target processing model is optimized again to obtain a second target processing model;

[0008] According to the second target processing model, the bioelectric signal to be measured is processed accordingly.

[0009] In combination with the first aspect, an embodiment of the present invention provides a first possible implementation of the first aspect, wherein, based on the separation processing of the current bioelectric signal, the first target processing model is optimized again to obtain the second target processing model, including:

[0010] Based on the first target processing model, a preset number of separation processes are performed on the current bioelectric signal collected in real time, and network model parameters corresponding to the target number of separation processing operations are determined, wherein the separation accuracy of the target number of separation processing operations meets the preset requirements;

[0011] The first target processing model is optimized according to the network model parameters to obtain a second target processing model.

[0012] In combination with the first aspect, an embodiment of the present invention provides a second possible implementation of the first aspect, wherein the step of performing a preset number of separation processes on the current bioelectric signal collected in real time based on the first target processing model and determining the network model parameters corresponding to the target number of separation processing operations includes:

[0013] The current bioelectric signal collected in real time is input into the first target processing model, and the following separation processing process is repeatedly performed until the current number of repetitions reaches a preset number of times:

[0014] According to the initial network model parameters corresponding to the first target processing model, the current bioelectric signal is separated and processed to obtain a first target separation signal and a first network model parameter; according to the first network model parameters, the first target processing model is updated to obtain a second target processing model, and the first target separation signal after the mixing process is input into the second target processing model for separation processing to obtain a second target separation signal and a second network model parameter corresponding to the current number; the second network model parameters are used as new initial network model parameters, the first target processing model is updated and the current bioelectric signal is separated and processed for the next number of times, wherein the next number is used as the new current number;

[0015] The second target separation signal is compared with the first target separation signal each time to determine the target number of separation processing operations in which the comparison result between the second target separation signal and the first target separation signal meets the preset requirements.

[0016] In combination with the first aspect, an embodiment of the present invention provides a third possible implementation manner of the first aspect, wherein the step of optimizing the first target processing model according to the network model parameters to obtain the second target processing model includes:

[0017] Performing a target processing operation on the first network model parameter and the second network model parameter corresponding to the target number of times to obtain a target network parameter, wherein the target processing operation includes average calculation and median calculation;

[0018] The first target processing model is optimized based on the target network parameters.

[0019] In combination with the first aspect, an embodiment of the present invention provides a fourth possible implementation of the first aspect, wherein the step of selecting a target audio separation processing model and optimizing the target audio separation processing model based on a bioelectric signal data set to obtain a first target processing model includes:

[0020] Determine a target audio separation processing model from the audio separation processing method;

[0021] Based on the artificial data set of bioelectric signals, the target audio separation processing model is optimized to obtain a first target processing model.

[0022] In combination with the first aspect, an embodiment of the present invention provides a fifth possible implementation of the first aspect, wherein the step of determining a target audio separation processing model from the audio separation processing method includes:

[0023] The target audio separation processing model is determined according to the processing capability of the time domain signal in each audio separation processing model and the memory usage of the model structure, wherein the target audio separation processing model includes a fully convolutional time domain audio separation network model.

[0024] In combination with the first aspect, an embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein the step of performing model architecture optimization processing on the target audio separation processing model based on an artificial data set of bioelectric signals to obtain a first target processing model includes:

[0025] While ensuring the processing performance of the artificial data set of bioelectric signals input into the target audio separation processing model, the first target processing model with optimized model architecture is obtained by reducing the number of network layers in the target audio separation processing model, the number of modules in each network layer, and the number of input values ​​of each module.

[0026] In a second aspect, an embodiment of the present invention further provides a bioelectric signal separation and noise reduction device, comprising:

[0027] A first optimization module selects a target audio separation processing model and optimizes the target audio separation processing model based on a bioelectric signal data set to obtain a first target processing model;

[0028] A second optimization module, based on the separation processing of the current bioelectric signal, further optimizes the first target processing model to obtain a second target processing model;

[0029] The processing module performs corresponding processing on the bioelectric signal to be measured according to the second target processing model.

[0030] In a third aspect, an embodiment provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any of the aforementioned implementation methods are implemented.

[0031] In a fourth aspect, an embodiment provides a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the steps of the method described in any one of the aforementioned implementations.

[0032] The embodiments of the present invention provide a method, device and electronic device for separating and denoising bioelectric signals, which determine a target audio separation and processing model from numerous audio separation and processing methods, and optimize the model structure for a bioelectric signal data set to obtain a first target processing model. The first target processing model is then optimized again based on the separation accuracy of the separation and processing of the current bioelectric signal according to the first target processing model to obtain a second target processing model. The optimized second target processing model can ensure that the separation of the bioelectric signal meets preset requirements, i.e., the bioelectric signal has anti-interference performance. Corresponding processing of the bioelectric signal using the second target processing model corresponding to the audio separation and processing field can obtain relatively accurate processing results.

[0033] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A flow chart of a bioelectric signal separation and noise reduction method provided by an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of comparing a speech signal and a bioelectric signal provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of optimizing a target audio separation processing model provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of functional modules of a bioelectric signal separation and noise reduction device provided by an embodiment of the present invention;

[0040] Figure 5 A schematic diagram of the hardware architecture of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] At present, under the vigorous development of science and technology, speech signal processing technology has also shown a hot development in recent years and has entered every aspect of our lives. Mixed speech separation is a technology that is familiar to us and can be classified as a speech signal processing algorithm.

[0043] The inventors have discovered that the audio separation processing method can be applied to the field of bioelectric signal diagnosis to solve the technical problem that bioelectric signals are susceptible to interference, such as separating the required bioelectric signals from the noise signals, thereby leaving the required bioelectric signals. However, the specific implementation method of the fusion and interaction of these two technologies is often overlooked.

[0044] Based on this, the embodiments of the present invention provide a method, device and electronic device for separating and denoising bioelectric signals, which, by applying the audio separation processing method to the field of separating and denoising bioelectric signals, alleviate the technical problem of inaccurate diagnostic results caused by interference with bioelectric signals.

[0045] To facilitate understanding of this embodiment, a bioelectric signal separation and noise reduction method disclosed in an embodiment of the present invention is first introduced in detail. This method can be applied to intelligent control devices such as controllers, host computers and servers.

[0046] Figure 1 A flow chart of a bioelectric signal separation and noise reduction method is provided for an embodiment of the present invention.

[0047] like Figure 1 As shown, the method may include the following steps:

[0048] Step S102, selecting a target audio separation processing model, and optimizing the target audio separation processing model based on the bioelectric signal data set to obtain a first target processing model.

[0049] Among them, the core points of the present invention include applying the audio separation processing algorithm to bioelectric signal separation and noise reduction, selecting a method suitable for bioelectric signal separation and noise reduction from a variety of speech algorithms, and optimizing the algorithm to facilitate better processing results for bioelectric signals.

[0050] Step S104, based on the separation processing of the current bioelectric signal, the first target processing model is optimized again to obtain a second target processing model.

[0051] Among them, in view of the fact that bioelectric signals are easily interfered with, the signal separation processing performance of the aforementioned optimized first target processing model is further optimized through the current bioelectric signals collected naturally, so that the optimized second target processing model can avoid the influence of interference on the bioelectric signals.

[0052] Step S106: Process the bioelectric signal to be measured accordingly according to the second target processing model.

[0053] Among them, the bioelectric signal to be tested can be processed in a variety of ways, including recognition, noise reduction, etc., and different audio separation processing methods can be used to optimize the second target processing model, and use this model to achieve better processing effects.

[0054] In a preferred embodiment of practical application, a target audio separation processing model is determined from numerous audio separation processing methods, and the model structure is optimized for a bioelectric signal data set to obtain a first target processing model. Then, based on the separation accuracy of the current bioelectric signal separated and processed by the first target processing model, the first target processing model is optimized again to obtain a second target processing model. The optimized second target processing model can ensure that the separation of the bioelectric signal meets preset requirements, i.e., the bioelectric signal has anti-interference performance. Corresponding processing of the bioelectric signal using the second target processing model corresponding to the audio separation processing field can obtain relatively accurate processing results.

[0055] As we all know, the audio separation processing algorithm is developed for speech signals, not for bioelectric signals, so the audio separation processing algorithm lacks specificity. This means that the algorithm that performs well in speech signal processing may not be suitable for bioelectric signal separation and noise reduction. For example, the speech signal processing environment is much more complicated than the electrocardiogram signal. Among them, background noise, the characteristics of each speaker's voice code, etc., require the audio separation processing algorithm to consider more problems, which inevitably increases the complexity of the audio separation processing algorithm. In actual applications, it also produces disadvantages such as excessive energy consumption caused by excessive calculations and large file memory usage, which limits the application of audio signals on wearable devices. At the same time, the complex algorithm also causes the speech algorithm to produce over-coupling when processing bioelectric signals, which has a very bad impact on the results.

[0056] At the same time, the inventors have found that the characteristic values ​​of bioelectric signals can be found in the time domain, which means that the algorithm specifically used to process bioelectric signals is designed under the premise that the time domain signal is used as the input signal. However, due to the particularity of speech signals, people cannot find the characteristic values ​​of speech signals in the time domain in most cases, but it is easier to find the characteristic values ​​of speech signals in the time-frequency domain, which means that the audio separation processing algorithm is designed under the premise that the time-frequency domain signal is used as the input signal in most cases. If the audio separation processing method needs to be transferred to bioelectric signal separation and noise reduction, the original bioelectric signal (time domain) needs to be transformed into the time-frequency domain through a short-time Fourier transform, and then processed by the audio separation processing algorithm, and then transformed back to the time domain through a short-time Fourier transform to obtain the processed bioelectric signal result. In other words, an algorithm that performs well in speech signal processing may not be suitable for bioelectric signal separation and noise reduction.

[0057] In some embodiments, in order to use the audio separation processing method to achieve better processing results for bioelectric signals, a suitable processing model can be selected from a variety of audio separation processing methods and the model can be adapted to bioelectric signal separation and noise reduction. As an optional embodiment, step S102 includes:

[0058] Step 1.1), determine the target audio separation processing model from the audio separation processing method.

[0059] Among them, the target audio separation processing model is determined according to the processing capability of the time domain signal in each audio separation processing model and the memory usage of the model structure.

[0060] It should be noted that among the numerous audio separation processing algorithms, we need to find an algorithm suitable for processing bioelectric signals. Considering the product requirements, we need an algorithm with a relatively simple structure, small file memory usage, and suitable for ultra-low power MCU. At the same time, with the development of technology, algorithms that can directly process time-domain voice signals are also being developed. Algorithms that directly process time-domain signals are also a plus point for us in selecting algorithms.

[0061] Here, the processing capability of the time domain signal can be understood as the ability to directly input the time domain signal into the audio separation processing model for calculation without the need for Fourier transform, and to ensure the correctness of the selected model through a large number of experimental verifications.

[0062] Among them, the target audio separation processing model can be a fully convolutional time-domain audio separation network model (convolutional time-domain audio separation network, Conv-TasNet).

[0063] Among the algorithms in Table 1, Conv-TasNet does not have the best signal-to-noise ratio, which means that Conv-TasNet does not perform the best in separating speech signals. However, we still chose this algorithm for bioelectric signal separation and noise reduction because Conv-TasNet has the smallest model size and can directly process time-domain speech signals. Finally, Conv-TasNet has a high number of citations in academia.

[0064] Table 1

[0065]

[0066]

[0067] Step 1.2), based on the artificial data set of bioelectric signals, the model architecture of the target audio separation processing model is optimized to obtain the first target processing model.

[0068] Among them, based on the defects of the audio separation processing algorithm itself, the algorithm structure needs to be optimized first. Without affecting the performance of the algorithm itself, the algorithm structure needs to be simplified, such as reducing the number of algorithm layers, reducing the number of Block modules in each layer, and reducing the input and output values ​​of each Block module. Of course, attention should also be paid to the selection of bioelectric signal feature values, and appropriate features should be selected to replace the feature values ​​of the audio separation processing algorithm.

[0069] Exemplarily, while ensuring the processing performance of the artificial data set of bioelectric signals input into the target audio separation processing model, the first target processing model after model architecture optimization is obtained by reducing the number of network layers in the target audio separation processing model, the number of modules in each network layer, and the number of input values ​​of each module.

[0070] like Figure 3 As shown in the figure, the Conv-TasNet algorithm diagram is used to illustrate the model architecture optimization. Figure 3 We can see that the algorithm repeats the operations of several layers of neural networks, so the algorithm structure can be optimized by reducing the number of repeated operations. Similarly, each layer also has many blocks modules to perform convolution processing on the signal. At this time, the number of blocks modules can also be reduced to optimize the algorithm structure. Finally, various operations need to be performed in each blocks module. The algorithm structure can be optimized by reducing the operation data in each blocks module.

[0071] In actual applications, when optimizing the algorithm structure, the algorithm structure can be simplified step by step through the program according to the characteristics of the bioelectric signal; for example, only one parameter of the algorithm structure is simplified each time. If the algorithm operation result develops in a good direction, the next parameter is optimized. If the algorithm operation result develops in a bad direction, the parameter is restored to its original state and the next parameter is optimized. This process is repeated until the program determines that the algorithm cannot be optimized again. When no better result is found, the model architecture optimization stops.

[0072] It should be noted that the selection of the audio separation processing algorithm that meets the requirements and can process bioelectric signals first depends on the processing operation category of the bioelectric signals. For example, if the processing operation category is recognition or noise reduction, the initial audio separation processing algorithm is selected, and then the memory structure size of the audio separation processing algorithm and the effect on audio signal processing are judged to find several algorithms suitable for processing bioelectric signals. Finally, the performance of these algorithms is verified and optimized through experiments until the most suitable algorithm is found.

[0073] On the basis of the aforementioned embodiments, the embodiments of the present invention find out an algorithm suitable for processing bioelectric signals from numerous audio separation processing algorithms through various experiments, and then determine the best model architecture parameters through optimization and verification of a large number of data sets.

[0074] After the model architecture parameters are optimized, there is another application challenge. The previous structural optimization verification experiments are based on artificial data sets of bioelectric signals. Artificial data sets have limitations. In order to cope with more challenges in actual situations, such as when using the audio separation processing algorithm for separation and noise reduction, bioelectric signals are easily interfered with, so it is not known which signal in the mixed signal is the desired original signal and which signal is the interference signal. This also causes the characteristic values ​​of bioelectric signals to be easily covered by interference signals. For example, the electrocardiogram signal of the fetus and the electrocardiogram signal of the mother are mixed together, and it is difficult to find the characteristic value of the electrocardiogram of the mother or the fetus.

[0075] In some embodiments, the optimal parameters of the audio separation processing algorithm for processing bioelectric signals can be found by adjusting parameters and optimizing parameters, so as to separate and process the interfered bioelectric signals. The aforementioned step S104 can also be implemented by the following steps, specifically including:

[0076] Step 2.1), based on the first target processing model, the current bioelectric signal collected in real time is separated and processed for a preset number of times, and the network model parameters corresponding to the target number of separation and processing operations are determined, and the separation accuracy of the target number of separation and processing operations meets the preset requirements.

[0077] Exemplarily, the current bioelectric signal collected in real time is input into the first target processing model, and the following separation processing process is repeatedly performed until the current number of repetitions reaches a preset number of times:

[0078] According to the initial network model parameters corresponding to the first target processing model, the current bioelectric signal is separated and processed to obtain the first target separation signal and the first network model parameters; according to the first network model parameters, the first target processing model is updated to obtain the second target processing model, and the first target separation signal after the mixing process is input into the second target processing model for separation processing to obtain the second target separation signal and the second network model parameters corresponding to the current number; the second network model parameters are used as the new initial network model parameters, the first target processing model is updated and the current bioelectric signal is separated and processed for the next number of times, wherein the next number is used as the new current number;

[0079] The second target separation signal is compared with the first target separation signal each time, and the separation processing operation of the target number of times that the comparison result of the second target separation signal and the first target separation signal meets the preset requirements is determined.

[0080] This parameter optimization method can quickly verify a large amount of data, thereby ensuring the reliability of the optimized second target processing model.

[0081] Step 2.2), optimize the first target processing model according to the network model parameters to obtain the second target processing model.

[0082] Here, the first network model parameter and the second network model parameter corresponding to the target number may be subjected to target processing operations to obtain target network parameters, and the first target processing model may be optimized based on the target network parameters. The target processing operations include average calculation and median calculation.

[0083] It should be noted that if Figure 2 As shown, speech signals and bioelectric signals are both one-dimensional signals, that is, their horizontal axis units are time, and their vertical axis units are intensity. Figure 2 The upper middle part is the ECG signal, and the lower part is the heart sound signal. In the process of signal collection, whether it is a voice signal or a bioelectric signal, the corresponding original signal is collected by a sensor, and then converted into a digital signal that can be recognized by the electronic device, and then converted into an array that can be accepted by the algorithm. Because voice signals and electrical signals are one-dimensional signals, the converted arrays are also one-dimensional arrays. Therefore, the embodiment of the present invention can use an audio separation processing algorithm to process bioelectric signals, thereby improving the performance of the product. Hybrid speech separation algorithm and speech noise reduction algorithm can be used to reduce the noise of bioelectric signals. Speech recognition algorithm can be used for intelligent diagnosis of bioelectric signals.

[0084] As an optional embodiment, because the ECG signal of the fetus is easily interfered by the ECG signal of the mother, the ECG signal of the fetus itself is very difficult to measure. An audio separation processing algorithm with a mixed voice separation function can be used to reduce the noise of the ECG signal of the fetus. First, the voice separation function is used to separate the ECG signal of the fetus from the ECG signal of the mother, so as to achieve the effect of noise reduction, and a clean and clear fetal ECG can be obtained. The audio separation processing algorithm can also be implemented by Conv-TasNet. This algorithm can directly process the time domain audio signal, and the memory occupied is not too large. After the input signal is converted into the array required by the algorithm, it is separated by a one-dimensional convolution layer. After separation, only the signal we need is left to achieve the effect of noise reduction. The separated signal is compared with the original signal to obtain an evaluation of the separation effect. During the evaluation process, the artificial data set signal can be used to optimize the architecture of the audio separation and processing algorithm model. The algorithm separates the fetal ECG signal, and then uses the naturally collected bioelectric signals to adjust the parameters and optimize the audio separation and processing algorithm model again. After that, the naturally collected bioelectric signals to be processed are placed into the audio separation and processing algorithm model after parameter adjustment for separation. The results are evaluated through expert assessment and AI judgment, and finally an experimental report is produced.

[0085] The embodiment of the present invention applies the audio separation processing algorithm to the separation and noise reduction of bioelectric signals. For example, first, in the process of signal acquisition and preprocessing, Fourier transform may not be used to convert the signal from the time domain to the time-frequency domain, and the signal can be directly entered into the algorithm from the time domain, thereby reducing operations and achieving the purpose of reducing energy consumption and reducing operation time. Secondly, an audio separation processing algorithm that meets the requirements and has a simple structure and small memory usage can be found and applied to the field of bioelectric signal separation and noise reduction. After that, the algorithm is optimized by adjusting parameters, and the optimal solution for processing bioelectric signals is achieved by continuously simplifying the algorithm structure. When extracting the eigenvalues ​​of bioelectric signals, a method combining artificial intelligence and experts is used to obtain more extensive verification.

[0086] like Figure 4 As shown, an embodiment of the present invention provides a bioelectric signal separation and noise reduction device, comprising:

[0087] A first optimization module selects a target audio separation processing model and optimizes the target audio separation processing model based on a bioelectric signal data set to obtain a first target processing model;

[0088] A second optimization module, based on the separation processing of the current bioelectric signal, further optimizes the first target processing model to obtain a second target processing model;

[0089] The processing module performs corresponding processing on the bioelectric signal to be measured according to the second target processing model.

[0090] Compared with the algorithm specifically used for separation and noise reduction of bioelectric signals, the speech algorithm of this application has the characteristics of more selectivity, rapid development and high maturity. At the same time, the frequency of speech signals is much higher than that of bioelectric signals, which means that the speech signal algorithm has to calculate more data. When developing these algorithms, the developers have noticed the importance of real-time performance to the implementation of the algorithm. Therefore, the speech signal processing algorithm is more suitable for separation and noise reduction of bioelectric signals of wearable products. It can be seen from the example of Conv-TasNet given in our previous embodiment that the audio separation processing algorithm can obtain results with better real-time performance and higher processing accuracy.

[0091] In some embodiments, the second optimization module is further specifically used to perform separation processing on the current bioelectric signal collected in real time a preset number of times based on the first target processing model, determine the network model parameters corresponding to the target number of separation processing operations, wherein the separation accuracy of the target number of separation processing operations meets the preset requirements; optimize the first target processing model according to the network model parameters to obtain the second target processing model.

[0092] In some embodiments, the second optimization module is also specifically used to input the current bioelectric signal collected in real time into the first target processing model, and repeatedly perform the following separation processing process until the current number of repetitions reaches a preset number: according to the initial network model parameters corresponding to the first target processing model, the current bioelectric signal is separated and processed to obtain the first target separation signal and the first network model parameters; according to the first network model parameters, the first target processing model is updated to obtain the second target processing model, and the first target separation signal after mixing is input into the second target processing model for separation processing to obtain the second target separation signal and the second network model parameters corresponding to the current number; using the second network model parameters as the new initial network model parameters, updating the first target processing model and performing the next number of separation processing on the current bioelectric signal, wherein the next number is used as the new current number; comparing the second target separation signal with the first target separation signal each time, and determining that the comparison result between the second target separation signal and the first target separation signal meets the preset requirements of the target number of separation processing operations.

[0093] In some embodiments, the second optimization module is further specifically used to perform target processing operations on the first network model parameters and the second network model parameters corresponding to the target times to obtain target network parameters, wherein the target processing operations include average calculation and median calculation; and optimize the first target processing model based on the target network parameters.

[0094] In some embodiments, the first optimization module is further specifically used to determine a target audio separation processing model from an audio separation processing method; based on an artificial data set of bioelectric signals, perform model architecture optimization processing on the target audio separation processing model to obtain a first target processing model.

[0095] In some embodiments, the first optimization module is further specifically used to determine the target audio separation processing model based on the processing capability of the time domain signal in each audio separation processing model and the memory usage of the model structure, wherein the target audio separation processing model includes a fully convolutional time domain audio separation network model.

[0096] In some embodiments, the first optimization module is further specifically used to obtain a first target processing model after model architecture optimization by reducing the number of network layers in the target audio separation processing model, the number of modules in each network layer, and the number of input values ​​of each module, while ensuring the processing performance of the artificial data set of bioelectric signals input into the target audio separation processing model.

[0097] An embodiment of the present invention provides an electronic device for implementing an electronic device. In this embodiment, the electronic device may be, but is not limited to, a personal computer (PC), a laptop computer, a monitoring device, a server, or other computer device with analysis and processing capabilities.

[0098] As an exemplary embodiment, see Figure 5 The electronic device 110 includes a communication interface 111, a processor 112, a memory 113 and a bus 114. The processor 112, the communication interface 111 and the memory 113 are connected via the bus 114. The memory 113 is used to store a computer program that supports the processor 112 to execute the method. The processor 112 is configured to execute the program stored in the memory 113.

[0099] The machine-readable storage medium mentioned in this article can be any electronic, magnetic, optical or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium can be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), any type of storage disk (such as CD, DVD, etc.), or similar storage medium, or a combination thereof.

[0100] The non-volatile medium may be a non-volatile memory, a flash memory, a storage drive (such as a hard drive), any type of storage disk (such as a CD, DVD, etc.), or a similar non-volatile storage medium, or a combination thereof.

[0101] It can be understood that the specific operation methods of each functional module in this embodiment can refer to the detailed description of the corresponding steps in the above method embodiment, and will not be repeated here.

[0102] The computer-readable storage medium provided in the embodiment of the present invention stores a computer program. When the computer program code is executed, the method described in any of the above embodiments can be implemented. For specific implementation, please refer to the method embodiment, which will not be described in detail here.

[0103] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0104] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0105] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0106] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.

Claims

1. A method for separating and denoising bioelectric signals, characterized in that: include: Selecting a target audio separation processing model, and optimizing the target audio separation processing model based on the bioelectric signal data set to obtain a first target processing model; The current bioelectric signal collected in real time is input into the first target processing model, and the following separation processing process is repeatedly performed until the current number of repetitions reaches a preset number of times: According to the initial network model parameters corresponding to the first target processing model, the current bioelectric signal is separated and processed to obtain a first target separation signal and a first network model parameter; according to the first network model parameters, the first target processing model is updated to obtain a second target processing model, and the first target separation signal after the mixing process is input into the second target processing model for separation processing to obtain a second target separation signal and a second network model parameter corresponding to the current number; the second network model parameters are used as new initial network model parameters, the first target processing model is updated and the current bioelectric signal is separated and processed for the next number of times, wherein the next number is used as the new current number; Comparing the second target separation signal with the first target separation signal each time, and determining the separation processing operation of the target number of times that the comparison result of the second target separation signal with the first target separation signal meets the preset requirements; The first target processing model is optimized according to the network model parameters corresponding to the target number of separation processing operations to obtain a second target processing model; and the bioelectric signal to be measured is processed accordingly according to the second target processing model.

2. The method according to claim 1, characterized in that The step of optimizing the first target processing model according to the network model parameters corresponding to the target number of separation processing operations to obtain the second target processing model comprises: Performing a target processing operation on the first network model parameter and the second network model parameter corresponding to the target number of times to obtain a target network parameter, wherein the target processing operation includes average calculation and median calculation; The first target processing model is optimized based on the target network parameters.

3. The method according to claim 1, characterized in that The step of selecting a target audio separation processing model and optimizing the target audio separation processing model based on the bioelectric signal data set to obtain a first target processing model includes: Determine a target audio separation processing model from the audio separation processing method; Based on the artificial data set of bioelectric signals, the target audio separation processing model is optimized to obtain a first target processing model.

4. The method according to claim 3, characterized in that The step of determining a target audio separation processing model from the audio separation processing method comprises: The target audio separation processing model is determined according to the processing capability of the time domain signal in each audio separation processing model and the memory usage of the model structure, wherein the target audio separation processing model includes a fully convolutional time domain audio separation network model.

5. The method according to claim 3, characterized in that: The step of performing model architecture optimization processing on the target audio separation processing model based on an artificial data set of bioelectric signals to obtain a first target processing model includes: While ensuring the processing performance of the artificial data set of bioelectric signals input into the target audio separation processing model, the first target processing model with optimized model architecture is obtained by reducing the number of network layers in the target audio separation processing model, the number of modules in each network layer, and the number of input values ​​of each module.

6. A bioelectric signal separation and noise reduction device, characterized in that: include: A first optimization module selects a target audio separation processing model and optimizes the target audio separation processing model based on a bioelectric signal data set to obtain a first target processing model; The second optimization module inputs the current bioelectric signal collected in real time into the first target processing model, and repeatedly performs the following separation processing process until the current number of repetitions reaches the preset number: according to the initial network model parameters corresponding to the first target processing model, the current bioelectric signal is separated and processed to obtain the first target separation signal and the first network model parameters; the first target processing model is updated according to the first network model parameters to obtain the second target processing model, and the first target separation signal after mixing is input into the second target processing model for separation processing to obtain the second target separation signal and the second network model parameters corresponding to the current number; the second network model parameters are used as the new initial network model parameters, the first target processing model is updated and the current bioelectric signal is separated and processed for the next number of times, wherein the next number is used as the new current number; the second target separation signal is compared with the first target separation signal each time to determine the target number of separation processing operations whose comparison results with the first target separation signal meet the preset requirements; the first target processing model is optimized according to the network model parameters corresponding to the target number of separation processing operations to obtain the second target processing model; The processing module performs corresponding processing on the bioelectric signal to be measured according to the second target processing model.

7. An electronic device, characterized in that: The method comprises a memory, a processor, and a program stored in the memory and capable of being run on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the program.

8. A computer-readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 5 is implemented.

Images