Heart sound data processing method and system based on time-frequency characteristic pattern

Abstract

The invention discloses a heart sound data processing method and system based on a time-frequency characteristic graph, and relates to the technical field of electric digital data processing. The heart sound data processing method based on the time-frequency characteristic graph comprises the following steps: steadily dividing heart sound, optimizing and identifying, and identifying and verifying noise. According to the method, feature stationary difference division is carried out on the time-frequency feature map obtained based on the heart sound signals to obtain the heart sound stationary difference feature result, then heart sound optimization processing is carried out in combination with the heart sound stationary difference feature result, the noise level recognition result is obtained, finally, the noise level recognition result is verified, and if verification succeeds, the noise level recognition result is recognized. If yes, feeding back the noise level recognition result, otherwise, updating the noise level recognition result, so that the effect of more accurately recognizing the noise of the heart sound signal is achieved, and the problem that the stability difference of the heart sound signal is not fully considered in the heart sound signal noise recognition process based on the time-frequency feature map in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular to a heart sound data processing method and system based on a time-frequency characteristic graph. Background Art

[0002] With the rapid development of the healthcare industry, especially the rise of personalized medicine and remote health monitoring, the analysis of heart sound signals is playing an increasingly important role in clinical diagnosis. Traditional methods for diagnosing heart disease, such as auscultation and electrocardiography, rely on the physician's experience and judgment, a subjective process that can easily lead to misdiagnosis or missed diagnosis. With advances in artificial intelligence and signal processing technologies, automated heart sound murmur recognition has become an important approach to improving diagnostic efficiency and accuracy. Heart sound signals are low-frequency biosignals that typically contain multiple components. Due to their diversity and complexity, accurately identifying murmurs and distinguishing healthy from diseased signals is challenging. Time-frequency feature maps convert heart sound signals into time and frequency domain representations, displaying the intensity and variations of different frequency components. This representation method facilitates observation of various characteristics of heart sounds, but it also faces some challenges. For example, time-frequency feature maps are generated using methods such as short-time Fourier transforms, which present a trade-off between time and frequency. Excessively long windows can lead to loss of temporal information, while too short windows can reduce frequency resolution, hindering murmur recognition. In heart sound signal analysis, especially in the application of time-frequency feature maps, obtaining a complete and high-quality heart sound dataset is a challenge. A lack of sufficient labeled data can lead to inadequate model training, which in turn affects recognition performance. Although identifying heart sound signal murmurs using time-frequency feature maps faces technical and data challenges, with the continuous development of signal processing technology, deep learning algorithms, wearable devices, and telemedicine, this field holds broad research and application prospects. With the accumulation of larger, higher-quality datasets, future research and technology will be able to overcome current difficulties, providing more accurate and efficient heart sound signal murmur recognition methods and promoting the prevention and diagnosis of early heart disease.

[0003] Existing time-frequency feature map technology can provide important information about heart health by collecting and analyzing heart sound signals. Murmur recognition is a crucial step in heart sound analysis, aiming to identify abnormal murmurs (such as heart disease, murmurs, or other abnormal sounds) from heart sound signals. Heart sound signals are typically collected using digital stethoscopes, wearable devices (such as chest straps, smartwatches, etc.), or electronic stethoscopes. After acquisition, heart sound signals typically require preprocessing to remove noise and interference and enhance signal quality. To extract time-frequency features from heart sound signals, techniques such as short-time Fourier transforms, wavelet transforms, or Mel-frequency cepstral coefficients are typically used to convert the heart sound signals into spectrograms in the time-frequency domain. With the advancement of deep learning, methods such as convolutional neural networks, long short-term memory networks, and recurrent neural networks have become mainstream for heart sound murmur recognition. Deep learning models can automatically learn features from spectrograms, avoiding the complexity of manual feature selection in traditional methods and offering greater expressiveness and robustness.

[0004] For example, the invention patent publication number CN111860246B discloses a data expansion method for heart sound signal classification for deep convolutional neural networks, including: preprocessing the original training heart sound data; converting the preprocessed one-dimensional heart sound signal into a two-dimensional MFSC feature map; multiplying the MFSC feature map by a mask function to perform random masking processing on its frequency domain and time domain; expanding and balancing the number of samples in each category to obtain a new training data set; and training a deep convolutional neural network with the new training data set.

[0005] For example, the invention patent publication number CN112036467B discloses an abnormal heart sound recognition method and device based on a multi-scale attention neural network, which includes: preprocessing the collected original heart sound signals and using the preprocessed heart sound signals as training samples; labeling the training samples for heart sound quality; training an abnormal heart sound recognition model based on the training samples and their labeled content; inputting the heart sound data to be detected into the trained abnormal heart sound recognition model to obtain a heart sound quality prediction result, and identifying abnormal heart sounds based on the heart sound quality prediction result.

[0006] However, in the process of implementing the technical solutions of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems: Time-frequency feature maps are used to extract and analyze various features in heart sound signals. When using time-frequency feature maps (such as short-time Fourier transform, etc.) to process heart sound data, it is often overlooked that heart sound signals may be non-stationary in a short period of time. In particular, in different cardiac cycles (such as different stages of atrial contraction and ventricular contraction), the frequency and amplitude of heart sound signals may change significantly, which is not enough to accurately reflect the actual heart sound signal characteristics.

[0007] In existing technologies, in actual heart sound signals, spectral overlap often occurs between noise (such as background noise, breathing sounds, and muscle vibrations) and heart sound signals. Due to the non-stationary nature of heart sound signals, their spectral characteristics change over time, making it more difficult to distinguish between heart sound signals and noise. If the non-stationarity of heart sound signals is not fully accounted for, the time-frequency feature graph may mistakenly identify noise as part of the heart sound, or fail to effectively separate the two, resulting in reduced recognition accuracy. Consequently, the problem of not fully considering the differences in stationarity of heart sound signals during heart sound signal-murmur recognition based on time-frequency feature graphs arises. Summary of the Invention

[0008] The present invention solves the problem in the prior art of not fully considering the differences in the stationarity of heart sound signals during the murmur recognition of heart sound signals based on time-frequency feature graphs by providing a heart sound data processing method and system based on time-frequency feature graphs, thereby achieving more accurate murmur recognition of heart sound signals.

[0009] The present invention provides a heart sound data processing method based on a time-frequency characteristic graph, comprising the following steps: S1, obtaining a heart sound signal within a preset time interval, performing feature stationary difference division on the time-frequency characteristic graph obtained based on the heart sound signal, and obtaining a heart sound stationary difference characteristic result; S2, performing heart sound optimization processing based on the heart sound stationary difference characteristic result to obtain a murmur level recognition result, wherein the heart sound optimization processing means optimizing the heart sound signal input in the murmur recognition process based on the heart sound stationary difference characteristic result to improve the accuracy of murmur recognition of the heart sound signal; S3, verifying the murmur level recognition result, and if the verification is successful, feeding back the murmur level recognition result; otherwise, updating the murmur level recognition result.

[0010] Furthermore, feature stationary difference division is performed on the time-frequency feature graph obtained based on the heart sound signal. The specific steps are as follows: obtaining feature division data of the heart sound signal in a preset time interval, the feature division data including signal autocorrelation value, signal stationary test root, signal frequency value, signal average frequency and average power spectrum density; performing difference averaging approximation operation on the signal autocorrelation value of the heart sound signal at the preset time interval at the preset acquisition and monitoring moment and the signal autocorrelation value at the left adjacent acquisition and monitoring moment, performing time domain stationary interaction processing on the signal stationary test root approximation operation result to obtain time domain stationarity. The time domain stationary interaction processing is used to describe the interaction between the signal autocorrelation value difference averaging approximation operation result and the signal stationary test root approximation operation result. ; After performing a difference analysis on the signal frequency value and the signal average frequency, the frequency domain smooth interaction processing is performed on the average power spectrum density approximation operation result to obtain the frequency domain smoothness. The frequency domain smooth interaction processing is used to describe the interaction between the signal frequency value difference analysis result and the average power spectrum density approximation operation result; the time domain smoothness, the frequency domain smoothness and the corresponding smoothness compensation amount are weighted and coupled to obtain the heart sound smooth division value. The heart sound smooth division value is used to quantitatively evaluate the smoothness of the central sound signal characteristics of the time-frequency feature graph; the smoothness compensation amount includes the time domain smoothness compensation amount and the frequency domain smoothness compensation amount; the heart sound smooth division value represents the quantitative data of the time domain smoothness and the frequency domain smoothness jointly evaluating the smoothness of the central sound signal characteristics of the time-frequency feature graph.

[0011] Furthermore, the heart sound stability difference characteristic result is obtained, and the specific process is as follows: according to the preset heart sound stability characteristic threshold range obtained from the preset database, it is judged whether the obtained heart sound stability division value is within the preset heart sound stability characteristic threshold range; if the heart sound stability division value is within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a stable heart sound signal; if the heart sound stability division value is not within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a non-stationary heart sound signal; the heart sound stability difference characteristic result includes a stable heart sound signal and a non-stationary heart sound signal.

[0012] Furthermore, the specific steps of performing heart sound optimization processing in combination with the heart sound smooth difference characteristic result are as follows: if the heart sound smooth difference characteristic result is a smooth heart sound signal, then performing smooth heart sound optimization processing on the smooth heart sound signal; smooth heart sound optimization processing means optimizing the smooth heart sound signal input based on the heart sound smooth difference characteristic result to improve the accuracy of the heart sound signal in the process of murmur identification; if the heart sound smooth difference characteristic result is a non-stationary heart sound signal, then performing non-stationary heart sound optimization processing on the non-stationary heart sound signal; non-stationary heart sound optimization processing means optimizing the non-stationary heart sound signal input based on the heart sound smooth difference characteristic result to improve the accuracy of the heart sound signal in the process of murmur identification; heart sound optimization processing includes smooth heart sound optimization processing and non-stationary heart sound optimization processing.

[0013] Furthermore, the specific process of performing smooth heart sound optimization processing on the smooth heart sound signal is as follows: the smooth heart sound signal is divided into short-time smooth heart sound signals in equal parts according to the smooth time window length, and a matching detection is performed based on the overlap between the short-time smooth heart sound signal and the reference known murmur signal to obtain an overlapping matching detection value, which is used to quantify the matching degree of the overlap between the short-time smooth heart sound signal and the reference known murmur signal; the smooth time window length represents the result of mapping the real-time heart sound smooth division value and the deviation degree of the heart sound smooth reference threshold into a smooth division length mapping set in a preset database, and the smooth division length mapping set represents the mapping relationship between the heart sound smooth division value and the deviation degree of the heart sound smooth reference threshold and the smooth time window length; if the deviation degree of the overlapping matching detection value from the reference overlapping matching threshold is within the preset allowable deviation range obtained from the preset database, then the murmur level recognition input priority of the short-time smooth heart sound signal is obtained based on the descending sorting of the input confidence score of the short-time smooth heart sound signal, and a preset number of short-time smooth heart sound signals are selected according to the murmur level recognition input priority. The short-time stationary heart sound signal is input into the established murmur level recognition model to obtain the murmur level recognition result; the input confidence score represents the result of mapping the deviation degree between the real-time overlap match detection value and the reference overlap match threshold into the input confidence mapping set in the preset database, and the input confidence mapping set represents the mapping relationship between the deviation degree between the overlap match detection value and the reference overlap match threshold and the input confidence score; if the deviation degree between the overlap match detection value and the reference overlap match threshold is not within the preset allowable deviation range obtained from the preset database, principal component analysis is performed based on the feature dimensionality reduction strength of the short-time stationary heart sound signal, and a preset number of short-time stationary heart sound signals after principal component analysis and dimensionality reduction are input into the established murmur level recognition model to obtain the murmur level recognition result; the feature dimensionality reduction strength represents the result of mapping the deviation degree between the real-time overlap match detection value and the reference overlap match threshold into the dimensionality reduction strength mapping set in the preset database, and the dimensionality reduction strength mapping set represents the mapping relationship between the deviation degree between the overlap match detection value and the reference overlap match threshold and the feature dimensionality reduction strength.

[0014] Furthermore, a matching detection is performed based on the overlap between the short-time stationary heart sound signal and the reference known murmur signal to obtain an overlapping matching detection value. The specific steps are as follows: obtaining overlapping detection quantified data between the short-time stationary heart sound signal and the reference known murmur signal, the overlapping detection quantified data including the detection mutual correlation coefficient, the detection similarity coefficient and the spectrum overlapping coefficient; the detection mutual correlation coefficient is used to describe the similarity between the short-time stationary heart sound signal and the reference known murmur signal in the time domain, the detection similarity coefficient is used to describe the structural similarity between the short-time stationary heart sound signal and the reference known murmur signal, and the spectrum overlapping coefficient is used to describe the similarity between the short-time stationary heart sound signal and the reference known murmur signal in the frequency domain; the overlapping detection quantified data and the corresponding detection compensation amount are weighted and coupled to obtain an overlapping matching detection value, the detection compensation amount including the cross-correlation detection compensation amount, the similarity detection compensation amount and the spectrum overlapping compensation amount; the overlapping matching detection value represents the quantified data of the overlapping detection quantified data on the overlapping matching detection between the short-time stationary heart sound signal and the reference known murmur signal.

[0015] Furthermore, the specific process of performing non-stationary heart sound optimization processing on non-stationary heart sound signals is as follows: judging whether the deviation degree between the heart sound stable division value and the heart sound stable reference threshold is greater than the stable deviation setting value obtained from the preset database; if the deviation degree between the heart sound stable division value and the heart sound stable reference threshold is greater than the stable deviation setting value obtained from the preset database, then the deviation degree between the heart sound stable division value and the stable deviation setting value is input into the heart sound sampling mapping set in the preset database for mapping to obtain the heart sound signal sampling frequency, and the heart sound signal is collected at the heart sound signal sampling frequency, and the heart sound sampling mapping set represents the mapping relationship between the deviation degree between the heart sound stable division value and the stable deviation setting value and the heart sound signal sampling frequency; if the deviation degree between the heart sound stable division value and the heart sound stable reference threshold is not greater than the stable deviation setting value obtained from the preset database, then the non-stationary heart sound signal is divided into short-term non-stationary heart sound signals according to the non-stationary time window length; the non-stationary time window length represents the deviation degree between the real-time heart sound stable division value and the heart sound stable reference threshold is input into the non-stationary division length mapping set in the preset database for mapping The non-stationary partition length mapping set represents the mapping relationship between the heart sound stationary partition value and the deviation degree of the heart sound stationary reference threshold and the non-stationary time window length; the signal energy value of the short-term non-stationary heart sound signal is obtained, and it is determined whether the signal energy value of the short-term non-stationary heart sound signal is greater than the reference signal energy threshold obtained from a preset database; if the signal energy value of the short-term non-stationary heart sound signal is greater than the reference signal energy threshold, the short-term non-stationary heart sound signal is marked as a valid heart sound signal; otherwise, the short-term non-stationary heart sound signal is marked as a waiting heart sound signal; it is determined whether the short-term non-stationary heart sound signals in adjacent non-stationary time windows of the valid heart sound signal are all valid heart sound signals, where the adjacent non-stationary time window length includes the left adjacent non-stationary time window and the right adjacent non-stationary time window of the valid heart sound signal; if the short-term non-stationary heart sound signals in the adjacent non-stationary time window length of the valid heart sound signal are all valid heart sound signals, the valid heart sound signal is marked as a valid priority heart sound signal; otherwise, it is not marked; the valid priority heart sound signal is input into the constructed murmur level recognition model to obtain a murmur level recognition result.

[0016] Furthermore, the murmur level recognition result is verified, and the specific process is as follows: if the heart sound stability difference characteristic result is a stable heart sound signal, a preset multiple number of short-term stable heart sound signals are input into the constructed murmur level recognition model to obtain a verified murmur level recognition result; if the heart sound stability difference characteristic result is a non-stationary heart sound signal, a preset multiple number of valid heart sound signals are input into the constructed murmur level recognition model to obtain a verified murmur level recognition result; determine whether the degree of conformity between the murmur level recognition result and the verified murmur level recognition result is greater than a preset conformity threshold obtained from a preset database; if the degree of conformity between the murmur level recognition result and the verified murmur level recognition result is greater than the preset conformity threshold obtained from the preset database, the verification is successful; otherwise, the murmur level recognition result is updated.

[0017] Furthermore, the specific steps for updating the murmur level recognition result are: verifying the murmur level recognition result; if the verification is still unsuccessful, verifying the output results in turn; if the verification is successful, updating the murmur level recognition result to the output result of the verification; when the number of input heart sound signals reaches the preset maximum number and the verification is still unsuccessful, updating the murmur level recognition result to recognition abnormality.

[0018] The present invention provides a heart sound data processing system based on a time-frequency feature graph, comprising: a heart sound steady division module, an optimization recognition processing module and a murmur recognition verification module; the heart sound steady division module is used to obtain a heart sound signal within a preset time interval, perform feature steady difference division on the time-frequency feature graph obtained based on the heart sound signal, and obtain a heart sound steady difference characteristic result; the optimization recognition processing module is used to perform heart sound optimization processing based on the heart sound steady difference characteristic result to obtain a murmur level recognition result, where the heart sound optimization processing means optimizing the heart sound signal input in the murmur recognition process based on the heart sound steady difference characteristic result to improve the accuracy of heart sound signal murmur recognition; the murmur recognition verification module is used to verify the murmur level recognition result, and if the verification is successful, the murmur level recognition result is fed back; otherwise, the murmur level recognition result is updated.

[0019] One or more technical solutions provided in the present invention have at least the following technical effects or advantages: 1. By performing feature stationary difference division on the time-frequency feature graph obtained based on the heart sound signal to obtain the heart sound stationary difference characteristic result, then combining the heart sound stationary difference characteristic result to perform heart sound optimization processing and obtain the murmur level recognition result, and finally verifying the murmur level recognition result. If the verification is successful, the murmur level recognition result is fed back; otherwise, the murmur level recognition result is updated, thereby realizing the stationary difference feature analysis of the heart sound signal and the optimization of the heart sound signal input in the murmur recognition process, thereby achieving more accurate murmur recognition of the heart sound signal, and effectively solving the problem in the prior art that the stationary difference of the heart sound signal is not fully considered in the murmur recognition process of the heart sound signal based on the time-frequency feature graph.

[0020] 2. By combining the results of the heart sound stability difference characteristics to perform heart sound optimization processing, if the heart sound stability difference characteristic result is a stable heart sound signal, the stable heart sound signal is subjected to stable heart sound optimization processing; if the heart sound stability difference characteristic result is a non-stationary heart sound signal, the non-stationary heart sound signal is subjected to non-stationary heart sound optimization processing, thereby achieving differential optimization of the heart sound signal input in murmur recognition by combining the difference in the stability of the heart sound signal, and further achieving improved input accuracy in the process of murmur recognition of the heart sound signal.

[0021] 3. By verifying the murmur level recognition result, if the verification is successful, the murmur level recognition result is updated to the output result for verification. Otherwise, the output results are verified in sequence. When the number of input heart sound signals reaches the preset maximum number and the verification is still not successful, the murmur level recognition result is updated to recognition abnormality, thereby achieving further verification of the murmur level recognition result, and further improving the reliability of the murmur level recognition result obtained by murmur recognition of heart sound signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of a heart sound data processing method based on a time-frequency feature map provided in an embodiment of the present application; Figure 2 A logic diagram of the steps of the heart sound data processing method based on the time-frequency feature map provided in an embodiment of the present application; Figure 3 This is a structural diagram of a heart sound data processing system based on a time-frequency feature graph provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiments of the present application provide a heart sound data processing method and system based on a time-frequency feature graph, thereby solving the problem in the prior art that the stationarity difference of the heart sound signal is not fully considered during the murmur recognition of the heart sound signal based on the time-frequency feature graph. The heart sound signal within a preset time interval is obtained, and then the time-frequency feature graph obtained based on the heart sound signal is divided into characteristic stationarity differences to obtain a heart sound stationarity difference characteristic result. Then, heart sound optimization processing is performed based on the heart sound stationarity difference characteristic result and a murmur level recognition result is obtained. Finally, the murmur level recognition result is verified. If the verification is successful, the murmur level recognition result is fed back. Otherwise, the murmur level recognition result is updated, thereby achieving more accurate murmur recognition of the heart sound signal.

[0024] The technical solution in the embodiment of the present application is to solve the problem that the stability difference of heart sound signals is not fully considered in the process of heart sound signal murmur recognition based on time-frequency feature maps. The overall idea is as follows: By performing feature stationary difference division on the time-frequency feature graph obtained based on the heart sound signal, then performing heart sound optimization processing based on the heart sound stationary difference characteristic results and obtaining the murmur level recognition result, and finally verifying the murmur level recognition result, a more accurate effect of heart sound signal murmur recognition is achieved.

[0025] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0026] like Figure 1 As shown, it is a flowchart of a heart sound data processing method based on a time-frequency feature graph provided by an embodiment of the present application, the method comprising the following steps: S1, obtaining a heart sound signal within a preset time interval, performing feature stationary difference division on the time-frequency feature graph obtained based on the heart sound signal, and obtaining a heart sound stationary difference characteristic result; S2, performing heart sound optimization processing based on the heart sound stationary difference characteristic result to obtain a murmur level recognition result, where the heart sound optimization processing means optimizing the heart sound signal input in the murmur recognition process based on the heart sound stationary difference characteristic result to improve the accuracy of murmur recognition of the heart sound signal; S3, verifying the murmur level recognition result. If the verification is successful, the murmur level recognition result is fed back; otherwise, the murmur level recognition result is updated.

[0027] In this embodiment, if Figure 2The figure shows a logic diagram of the steps of a heart sound data processing method based on a time-frequency feature map provided by an embodiment of the present application. Heart sound signals reflect cardiac activity, and cardiac activity itself exhibits different physiological and pathological manifestations at different time points. For example, the heart's operating states during systole and diastole vary, and the characteristics of heart sounds also change accordingly. Furthermore, pathological conditions such as heart disease and valvular heart disease can alter the heart's rhythm, blood flow, and the vibration characteristics of cardiac tissue, leading to abnormal heart sound waveforms and, in turn, signal non-stationarity. Furthermore, heart sounds include not only normal heart sounds (such as the first and second heart sounds) but also murmurs. Murmurs typically occur at specific times and vary with the condition. For example, murmurs caused by valvular heart disease often overlap with normal heart sounds, and their frequency and amplitude vary with the heart's activity cycle, leading to signal non-stationarity. Murmurs can also arise from environmental noise, equipment errors, patient movement, or changes in body position. The inherent characteristics of heart sound signals (such as the overlap of the heart sound and murmur spectrum) can also make noise identification more difficult.

[0028] In heart sound signal processing, the difference in the stationarity of heart sounds is crucial for correct signal processing and murmur identification. This difference directly affects the signal's feature extraction, model training, and final recognition results. Ignoring this difference may affect the accuracy of feature extraction, model training, and recognition results. To address this issue, the algorithm in this application optimizes the heart sound signal input during murmur recognition based on the analysis of the stationarity difference characteristics of heart sound signals. This effectively improves the accuracy of heart sound signal processing and murmur recognition, thereby providing a more reliable auxiliary diagnostic tool for clinical use.

[0029] Furthermore, the time-frequency feature map obtained based on the heart sound signal is divided into feature stationary differences. The specific steps are as follows: A1. Obtain feature division data of the intraocular sound signal in a preset time interval. The feature division data includes signal autocorrelation value, signal stationarity test root, signal frequency value, signal average frequency, and average power spectral density. Specifically, obtain the signal autocorrelation value through the autocorrelation function in the NumPy and SciPy libraries in Python; obtain the signal stationarity test root through the statsmodels library in Python (adfuller function performs ADF test); obtain the signal frequency value through numpy.fft in Python, and perform statistical analysis on the signal frequency value through the AVERAGE function in Excel to obtain the signal average frequency; obtain the signal power spectral density through using pwelch() in MATLAB, and perform statistical analysis on the signal power spectral density through the AVERAGE function in Excel to obtain the average power spectral density; the signal frequency value, signal average frequency, and average power spectral density are all denormalized.

[0030] Among them, before the design of the heart sound data processing method based on the time-frequency characteristic diagram provided in the present application, a preset database for storing various set data is established, which includes but is not limited to autocorrelation values, stationarity test roots, power spectrum density, etc., and various numerical values are directly set by technical personnel; for example, the minimum value of the difference operation between the autocorrelation value of the collected heart sound signal at the preset historical acquisition and monitoring moment and the signal at the left adjacent historical acquisition and monitoring moment represents the reference minimum autocorrelation difference; the reference stationarity test root is represented by the result of summing and averaging the stationarity test roots of the collected historical signals; the reference power spectrum density is represented by the result of summing and averaging the power spectrum density of the collected historical signals.

[0031] A2, after performing a difference averaging approximation operation on the autocorrelation value of the intracardiac signal at the preset acquisition and monitoring time and the signal at the left adjacent acquisition and monitoring time within the preset time interval, the root approximation operation result of the signal stationarity test is subjected to time domain stationarity interactive processing to obtain time domain stationarity. Time domain stationarity interactive processing is used to describe the interaction between the result of the difference averaging approximation operation of the signal autocorrelation value and the result of the root approximation operation of the signal stationarity test, specifically referring to the process of coupling the result of the difference averaging approximation operation of the signal autocorrelation value and the result of the root approximation operation of the signal stationarity test to obtain time domain stationarity. The specific restricted expression of time domain stationarity is; Where, represents the time domain stationarity, t represents the acquisition and monitoring time of the heart sound signal in the preset time interval, t=1,2,…,T, T represents the total number of acquisition and monitoring time of the heart sound signal in the preset time interval (T≥2), ZR(t) represents the signal autocorrelation value of the heart sound signal at the t-th acquisition and monitoring time in the preset time interval, ZR(t-1) represents the signal autocorrelation value of the heart sound signal at the t-1-th acquisition and monitoring time in the preset time interval, MIN zc Denotes the reference minimum autocorrelation difference, DG denotes the signal stationarity test root, and △DG denotes the reference stationarity test root.

[0032] A3, after performing a difference analysis between the signal frequency value and the signal average frequency, performs frequency domain stationary interaction processing on the average power spectrum density approximation operation result to obtain frequency domain stationarity. Frequency domain stationary interaction processing is used to describe the interaction between the signal frequency value difference analysis result and the average power spectrum density approximation operation result. Specifically, it refers to the process of coupling the signal frequency value difference analysis result and the average power spectrum density approximation operation result to obtain frequency domain stationarity. The specific restricted expression of frequency domain stationarity is: Where, represents the frequency domain stationarity, f(t) represents the signal frequency value of the tth acquisition monitoring moment of the heart sound signal within the preset time interval, P f Indicates the average frequency of the intraocular signal within a preset time interval, P GP It represents the average power spectrum density of the intraocular sound signal in the preset time interval, and △GP represents the reference power spectrum density.

[0033] A4, the time domain stationarity (i.e. ), frequency domain stability (i.e. ) is coupled with the corresponding steady compensation amount after weighted operation to obtain the steady division value of heart sound. The steady compensation amount includes the time domain steady compensation amount (i.e. ) and frequency domain stationary compensation (i.e. ), the time domain smoothness compensation amount and the frequency domain smoothness compensation amount are used to describe the degree of influence of time domain smoothness and frequency domain smoothness on the heart sound smoothness division value, respectively. The sum of the time domain smoothness compensation amount and the frequency domain smoothness compensation amount is 1. They are obtained from a preset database. For example, the real-time time domain smoothness and frequency domain smoothness are input into the preset mapping set of time domain smoothness, frequency domain smoothness and their corresponding compensation amounts in the database to obtain the time domain smoothness compensation amount and the frequency domain smoothness compensation amount. The specific restricted expression of the heart sound smoothness division value is: Where, Indicates the stable division value of heart sounds, represents the time domain steady compensation, Indicates the frequency domain stationary compensation amount.

[0034] In summary, the heart sound stable division value represents the quantitative data of the time domain stableness and frequency domain stableness jointly evaluating the stableness of the central sound signal characteristics of the time-frequency feature graph. The time domain stableness and frequency domain stableness in the heart sound stable division value jointly affect the stableness evaluation of the central sound signal characteristics of the time-frequency feature graph. Specifically, as the time domain stableness and frequency domain stableness increase, the heart sound stable division value increases accordingly.

[0035] In this embodiment, the heart sound stationary segmentation value is used to quantitatively assess the stationarity of the heart sound signal characteristics in the time-frequency feature map. This is achieved through quantitative analysis of the correlation between various parameters. These parameters influence each other and do not exist independently. For example, the signal autocorrelation value reflects the correlation of the heart sound signal at different time delays and is used to describe the temporal structure of the heart sound signal. Analysis of the differences in the signal autocorrelation values reveals whether the heart sound signal exhibits periodic or trending changes. The signal stationarity test is used to determine whether the statistical characteristics of the heart sound signal change over time. As the deviation between the difference between the signal autocorrelation values and the reference minimum autocorrelation difference decreases, the heart sound signal's temporal characteristics are relatively stable. This means that the deviation between the signal's stationarity test root and the reference stationarity test root also decreases, leading to an increase in time-domain stationarity. Furthermore, as the signal's frequency difference analysis results increase, indicating an increase in the frequency fluctuation of the heart sound signal, the frequency range of the heart sound signal's spectrum increases, meaning that the power spectrum of the signal may have a more pronounced energy distribution in the high-frequency range. This means that the deviation between the heart sound signal's average power spectral density and the reference power spectral density increases, leading to a decrease in frequency-domain stationarity. Furthermore, time-domain stationarity and frequency-domain stationarity are interrelated. A heart sound signal exhibiting high temporal stationarity generally indicates relatively stable frequency characteristics. Conversely, if a signal exhibits significant temporal variations (e.g., large autocorrelation value fluctuations), its frequency fluctuations and power spectral density in the frequency domain may also be unstable, resulting in low frequency-domain stationarity. In other words, as time-domain stationarity increases, frequency-domain stationarity also increases, leading to an increase in the heart sound stationary classification value.

[0036] By quantitatively evaluating the characteristic smoothness of the heart sound signal in the time-frequency feature graph, a more accurate evaluation of the characteristic smoothness of the heart sound signal is achieved, thereby achieving a more accurate division of the differences in the smoothness of the heart sound signal in the recognition of heart sound signal murmurs.

[0037] Furthermore, the heart sound stability difference characteristic results are obtained. The specific process is as follows: B1. Based on the preset heart sound stability characteristic threshold range obtained from the preset database, determine whether the obtained heart sound stability classification value is within the preset heart sound stability characteristic threshold range; the preset heart sound stability characteristic threshold range is set by professionals according to standards in the field.

[0038] B2. If the heart sound stability classification value is within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a stable heart sound signal.

[0039] B3. If the heart sound stability classification value is not within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a non-stationary heart sound signal.

[0040] In this embodiment, the heart sound stability difference characteristic results include a stable heart sound signal and a non-stationary heart sound signal. By combining a preset heart sound stability characteristic threshold range to judge the characteristic stability of the heart sound signal in the time-frequency feature graph, the heart sound stability difference characteristic results are obtained after the characteristic stability difference of the heart sound signal is divided and judged, thereby improving the reliability of the processing of the heart sound signal characteristic difference in the process of heart sound signal murmur recognition.

[0041] Furthermore, the specific steps for optimizing the heart sound processing based on the heart sound stationary difference characteristic results are as follows: C1. If the heart sound smooth difference characteristic result is a smooth heart sound signal, the smooth heart sound signal is subjected to smooth heart sound optimization processing; the smooth heart sound optimization processing means optimizing the smooth heart sound signal input based on the heart sound smooth difference characteristic result to improve the accuracy of the heart sound signal in the process of murmur recognition.

[0042] It should be added that the specific process of performing steady heart sound optimization processing on steady heart sound signals is as follows: C11, the stable heart sound signal is divided into short-term stable heart sound signals according to the stable time window length, and a matching detection is performed based on the overlap between the short-term stable heart sound signal and the reference known noise signal to obtain an overlapping matching detection value.

[0043] The overlap detection is performed based on the overlap between the short-term steady heart sound signal and the reference known murmur signal to obtain the overlap match detection value. The specific steps are as follows: C111. Obtain overlap detection quantitative data between the short-term stationary heart sound signal and the reference known murmur signal. The overlap detection quantitative data includes the detection mutual correlation coefficient, the detection similarity coefficient, and the spectrum overlap coefficient. The detection mutual correlation coefficient is obtained using the xcorr function in MATLAB; the detection similarity coefficient is obtained using the cosine function in MATLAB; and the spectrum overlap coefficient is obtained using the pwelch function in MATLAB.

[0044] Specifically, the detection cross-correlation coefficient is used to describe the similarity between the short-time stationary heart sound signal and the reference known noise signal in the time domain, the detection similarity coefficient is used to describe the structural similarity between the short-time stationary heart sound signal and the reference known noise signal, and the spectrum overlap coefficient is used to describe the similarity between the short-time stationary heart sound signal and the reference known noise signal in the frequency domain.

[0045] C112, weighted operation is performed on the overlap detection quantized data and the corresponding detection compensation amount, and then coupled to obtain the overlap matching detection value. The specific restricted expression of the overlap matching detection value is: Where, represents the overlap matching detection value, δ1 represents the cross-correlation detection compensation amount, δ2 represents the similarity detection compensation amount, δ3 represents the spectrum overlap compensation amount, represents the detection correlation coefficient, Represents the detection similarity coefficient, Represents the spectrum overlapping coefficient.

[0046] Among them, the detection compensation amount includes the mutual correlation detection compensation amount, the similarity detection compensation amount and the spectrum overlap compensation amount; the real-time detection mutual correlation coefficient, the detection similarity coefficient and the spectrum overlap coefficient are input into the mapping set of the preset detection mutual correlation coefficient, the detection similarity coefficient and the spectrum overlap coefficient and their corresponding compensation amounts in the database to obtain the mutual correlation detection compensation amount, the similarity detection compensation amount and the spectrum overlap compensation amount.

[0047] The overlap match detection value represents the quantitative data of the overlap detection quantification data for the overlap match detection of the short-term stationary heart sound signal and the reference known murmur signal. It is used to quantify the degree of overlap match between the short-term stationary heart sound signal and the reference known murmur signal. The overlap match detection value includes multiple parameters. Specifically, as the detection cross-correlation coefficient, detection similarity coefficient, and spectral overlap coefficient increase, the overlap match detection value increases, indicating an increase in the degree of overlap match between the short-term stationary heart sound signal and the reference known murmur signal. Furthermore, each parameter in the overlap match detection value is not independent. For example, the detection cross-correlation coefficient describes the similarity between the short-term stationary heart sound signal and the reference known murmur signal in the time domain. If the signals are well aligned in the time domain, that is, the detection cross-correlation coefficient increases, then their morphology and structure are likely to be highly similar, that is, the detection similarity coefficient also increases. Meanwhile, the spectral overlap coefficient describes the similarity between the short-term stationary heart sound signal and the reference known murmur signal in the frequency domain. When two signals have strong overlap in the frequency domain, their time domain waveforms are generally similar, especially in periodic signals. Therefore, the spectrum overlap coefficient is also interrelated with the detection cross-correlation coefficient and the detection similarity coefficient, which together quantify the matching degree of the overlap between the detected short-term stable heart sound signal and the reference known murmur signal.

[0048] The smooth time window length represents the result of mapping the real-time heart sound smooth division value and the degree of deviation from the heart sound smooth reference threshold into the smooth division length mapping set in the preset database. The smooth division length mapping set represents the mapping relationship between the heart sound smooth division value and the degree of deviation from the heart sound smooth reference threshold and the smooth time window length.

[0049] C12. If the degree of deviation between the overlap match detection value and the reference overlap match threshold is within a preset allowable deviation range obtained from a preset database, the murmur level identification input priority of the short-term steady heart sound signals is determined based on the descending order of the input confidence scores of the short-term steady heart sound signals. A preset number of short-term steady heart sound signals are input into the established murmur level identification model based on the murmur level identification input priority to obtain a murmur level identification result. The reference overlap match threshold is the sum and average of the collected historical overlap match detection values. The preset allowable deviation range is set by professionals based on field standards, for example, the preset allowable deviation range is set to 3.0 to 5.0.

[0050] The input confidence score involved represents the result of mapping the degree of deviation between the real-time overlap match detection value and the reference overlap match threshold into the input confidence mapping set in the preset database. The input confidence mapping set represents the mapping relationship between the degree of deviation between the overlap match detection value and the reference overlap match threshold and the input confidence score, wherein, as the degree of deviation between the overlap match detection value and the reference overlap match threshold increases, the input confidence score decreases accordingly.

[0051] C13. If the degree of deviation between the overlapping match detection value and the reference overlapping match threshold is not within the preset allowable deviation range obtained from the preset database, principal component analysis dimensionality reduction is performed based on the characteristic dimensionality reduction strength of the short-time stable heart sound signal, and a preset number of short-time stable heart sound signals after principal component analysis dimensionality reduction are input into the constructed murmur level recognition model to obtain the murmur level recognition result.

[0052] The feature dimensionality reduction strength involved represents the result of mapping the degree of deviation between the real-time overlapping matching detection value and the reference overlapping matching threshold into the dimensionality reduction strength mapping set in the preset database. The dimensionality reduction strength mapping set represents the mapping relationship between the degree of deviation between the overlapping matching detection value and the reference overlapping matching threshold and the feature dimensionality reduction strength, wherein as the degree of deviation between the overlapping matching detection value and the reference overlapping matching threshold increases, the feature dimensionality reduction strength increases accordingly.

[0053] C2. If the result of the heart sound stationary difference characteristic is a non-stationary heart sound signal, non-stationary heart sound optimization processing is performed on the non-stationary heart sound signal; non-stationary heart sound optimization processing means optimizing the non-stationary heart sound signal input based on the heart sound stationary difference characteristic result to improve the accuracy of the heart sound signal in the process of murmur recognition.

[0054] It should be added that the specific process of performing non-stationary heart sound optimization processing on non-stationary heart sound signals is as follows: C21, determines whether the degree of deviation between the heart sound smooth division value and the heart sound smooth reference threshold is greater than the smooth deviation setting value obtained from the preset database; the degree of deviation between the heart sound smooth division value and the heart sound smooth reference threshold is the result of taking the absolute value of the difference between the heart sound smooth division value and the heart sound smooth reference threshold; the heart sound smooth reference threshold and the smooth deviation setting value are obtained from the preset database, for example, the heart sound smooth reference threshold is set to the maximum value of the preset heart sound smooth characteristic threshold range, and the smooth deviation setting value is the maximum value of the degree of deviation between the collected historical heart sound smooth division value and the heart sound smooth reference threshold.

[0055] C22. If the degree of deviation between the heart sound stable division value and the heart sound stable reference threshold is greater than the stable deviation setting value obtained from the preset database, the degree of deviation between the heart sound stable division value and the stable deviation setting value is input into the heart sound sampling mapping set in the preset database for mapping to obtain the heart sound signal sampling frequency, and the heart sound signal is collected at the heart sound signal sampling frequency. The heart sound sampling mapping set represents the mapping relationship between the degree of deviation between the heart sound stable division value and the stable deviation setting value and the heart sound signal sampling frequency, wherein as the degree of deviation between the heart sound stable division value and the stable deviation setting value increases, the heart sound signal sampling frequency increases accordingly.

[0056] C23, if the deviation between the heart sound stability division value and the heart sound stability reference threshold is not greater than the stability deviation setting value obtained from the preset database, the non-stationary heart sound signal is divided into short-term non-stationary heart sound signals according to the non-stationary time window length.

[0057] The non-stationary time window length involved represents the result of mapping the degree of deviation between the real-time heart sound stable division value and the heart sound stable reference threshold into the non-stationary division length mapping set in the preset database. The non-stationary division length mapping set represents the mapping relationship between the heart sound stable division value and the degree of deviation between the heart sound stable reference threshold and the non-stationary time window length, wherein as the degree of deviation between the heart sound stable division value and the heart sound stable reference threshold increases, the non-stationary time window length decreases accordingly.

[0058] C24 obtains the signal energy value of the short-term non-stationary heart sound signal and determines whether the signal energy value of the short-term non-stationary heart sound signal is greater than a reference signal energy threshold obtained from a preset database. If the signal energy value of the short-term non-stationary heart sound signal is greater than the reference signal energy threshold, the short-term non-stationary heart sound signal is marked as a valid heart sound signal; otherwise, the short-term non-stationary heart sound signal is marked as a waiting heart sound signal. The signal energy value is obtained using a signal processing tool (such as LabVIEW), and the reference signal energy threshold is represented by summing and averaging the collected historical signal energy values.

[0059] C25, determines whether the short-term non-stationary heart sound signals in the adjacent non-stationary time windows of the valid heart sound signal are all valid heart sound signals, where the length of the adjacent non-stationary time window includes the left adjacent non-stationary time window and the right adjacent non-stationary time window of the valid heart sound signal; if the short-term non-stationary heart sound signals in the length of the adjacent non-stationary time window of the valid heart sound signal are all valid heart sound signals, then the valid heart sound signal is marked as a valid priority heart sound signal; otherwise, no marking is performed.

[0060] C26, inputting the effective priority heart sound signal into the constructed murmur level recognition model to obtain the murmur level recognition result.

[0061] In this embodiment, the murmur level recognition model incorporates three levels of attention: spatial attention, channel attention, and frame attention. Spatial attention is first embedded within each scale branch of the multi-scale feature extraction network. Its purpose is to enable the model to adaptively focus on highly discriminative regions in the mel-spectrogram, assigning higher weights to these regions when extracting murmurs, reducing interference from irrelevant regions and improving feature discriminability. Next, features at each scale are concatenated channel by channel and then aggregated with different weights using channel attention to determine the murmur level for a single cardiac cycle. This aggregation method can, to a certain extent, perform scale selection and more accurately extract murmur features from the mel-spectrogram. Finally, the frame attention mechanism adaptively determines the aggregation coefficient when aggregating features across cardiac cycles based on the complementary and redundant properties between cardiac cycles, reducing the impact of differences between cardiac cycles on murmur features and improving feature invariance. The three levels of attention mechanism are closely integrated; heart sound optimization processing includes steady heart sound optimization processing and non-steady heart sound optimization processing; by performing corresponding heart sound optimization processing when the heart sound steady difference characteristic result is a steady heart sound signal or a non-steady heart sound signal, the differential optimization processing of the heart sound signal input for murmur recognition is achieved by combining the steady difference of the heart sound signal, thereby achieving an improvement in the input accuracy of the heart sound signal during murmur recognition.

[0062] Furthermore, the noise level recognition results are verified. The specific process is as follows: D1. If the result of the heart sound stability difference characteristic is a stable heart sound signal, a preset multiple number of short-term stable heart sound signals is input into the constructed murmur level recognition model to obtain a verified murmur level recognition result; wherein, the preset multiple number is set by professionals according to standards in the field, for example, the preset multiple number is set to one times the number of input short-term stable heart sound signals.

[0063] D2. If the heart sound stationary difference characteristic result is a non-stationary heart sound signal, a preset multiple number of valid heart sound signals are input into the constructed murmur level recognition model to obtain a verified murmur level recognition result.

[0064] D3: Determine whether the degree of consistency between the noise level recognition result and the verification noise level recognition result is greater than a preset consistency threshold obtained from a preset database. If the degree of consistency between the noise level recognition result and the verification noise level recognition result is greater than the preset consistency threshold obtained from the preset database, the verification is successful; otherwise, the noise level recognition result is updated. The consistency is the ratio of the total number of times the noise level recognition result and the verification noise level recognition result are identical to each other to the total number of times the noise level recognition result is verified. The preset consistency threshold is set by professionals based on field standards, for example, the preset consistency threshold is set to 0.8.

[0065] It should be added that the specific steps for updating the noise level recognition results are as follows: D31, verifying the noise level recognition result. If the verification is still unsuccessful, verifying the output results in sequence. If the verification is successful, updating the noise level recognition result to the output result for verification.

[0066] D32: If the number of input heart sound signals reaches the preset maximum number and verification is still unsuccessful, the murmur level recognition result is updated to abnormal. The preset maximum number is set by professionals based on standards in the field. For example, the preset maximum number is set to five times the number of input heart sound signals.

[0067] In this embodiment, the noise level recognition result is verified and updated by combining the conformity of the noise level recognition result with the verification noise level recognition result and a preset conformity threshold, thereby realizing the verification and updating of the noise level recognition result obtained in the process of noise recognition of heart sound signals, thereby realizing the improvement of the reliability of the noise level recognition result obtained in the process of noise recognition of heart sound signals.

[0068] like Figure 3As shown, it is a structural schematic diagram of a heart sound data processing system based on a time-frequency feature graph provided in an embodiment of the present application. The heart sound data processing system based on a time-frequency feature graph provided in an embodiment of the present application includes: a heart sound steady division module, an optimization recognition processing module and a murmur recognition verification module; the heart sound steady division module is used to obtain heart sound signals within a preset time interval, perform feature steady difference division on the time-frequency feature graph obtained based on the heart sound signal, and obtain a heart sound steady difference characteristic result; the optimization recognition processing module is used to perform heart sound optimization processing based on the heart sound steady difference characteristic result to obtain a murmur level recognition result, where the heart sound optimization processing means optimizing the heart sound signal input in the murmur recognition process based on the heart sound steady difference characteristic result to improve the accuracy of murmur recognition of the heart sound signal; the murmur recognition verification module is used to verify the murmur level recognition result. If the verification is successful, the murmur level recognition result is fed back; otherwise, the murmur level recognition result is updated.

[0069] In summary, the embodiment of the present application performs feature stationary difference division on the time-frequency feature graph obtained based on the heart sound signal to obtain the heart sound stationary difference characteristic result, then performs heart sound optimization processing based on the heart sound stationary difference characteristic result and obtains the murmur level recognition result, and finally verifies the murmur level recognition result. If the verification is successful, the murmur level recognition result is fed back; otherwise, the murmur level recognition result is updated, thereby realizing the analysis of the stationary difference characteristics of the heart sound signal and the optimization of the heart sound signal input in the murmur recognition process, thereby realizing more accurate murmur recognition of the heart sound signal, and effectively solving the problem in the prior art that the stationary difference of the heart sound signal is not fully considered in the process of murmur recognition of the heart sound signal based on the time-frequency feature graph.

[0070] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0071] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A heart sound data processing method based on a time-frequency feature map, characterized in that: The following steps are involved: S1, obtaining a heart sound signal within a preset time interval, performing feature stationary difference division on a time-frequency feature map obtained based on the heart sound signal, and obtaining a heart sound stationary difference characteristic result; S2, performing heart sound optimization processing based on the heart sound stationary difference characteristic result to obtain a murmur level recognition result, wherein the heart sound optimization processing is to optimize the heart sound signal input in the murmur recognition process based on the heart sound stationary difference characteristic result to improve the accuracy of murmur recognition of the heart sound signal; S3, verifying the noise level recognition result. If the verification is successful, the noise level recognition result is fed back. Otherwise, the noise level recognition result is updated.

2. The heart sound data processing method based on the time-frequency feature map according to claim 1, characterized in that: The specific steps of performing feature stationary difference division on the time-frequency feature graph obtained based on the heart sound signal are as follows: Acquire characteristic division data of an intracardiac sound signal within a preset time interval, wherein the characteristic division data includes a signal autocorrelation value, a signal stationarity test root, a signal frequency value, a signal average frequency, and an average power spectrum density; After performing a difference averaging approximation operation on the autocorrelation value of the intracardiac signal at a preset acquisition and monitoring time and the signal at the left adjacent acquisition and monitoring time within a preset time interval, performing time domain stationary interaction processing on the signal stationary test root approximation operation result to obtain the time domain stationarity, the time domain stationary interaction processing is used to describe the interaction between the signal autocorrelation value difference averaging approximation operation result and the signal stationary test root approximation operation result; After performing a difference analysis on the signal frequency value and the signal average frequency, a frequency domain stationary interaction process is performed on the average power spectrum density approximation operation result to obtain the frequency domain stationarity. The frequency domain stationary interaction process is used to describe the interaction between the signal frequency value difference analysis result and the average power spectrum density approximation operation result; The time domain stationarity, the frequency domain stationarity and the corresponding stationarity compensation amount are weighted and coupled to obtain a heart sound stationarity division value, which is used to quantitatively evaluate the stationarity of the heart sound signal feature in the time-frequency feature graph; The said steady compensation amount includes a time domain steady compensation amount and a frequency domain steady compensation amount; The heart sound stability classification value represents quantitative data of the time domain stability and the frequency domain stability jointly evaluating the stability of the heart sound signal characteristics of the time-frequency feature graph.

3. The heart sound data processing method based on the time-frequency feature map according to claim 2, characterized in that: The specific process of obtaining the heart sound stability difference characteristic result is as follows: According to the preset heart sound stability characteristic threshold range obtained from the preset database, determining whether the obtained heart sound stability classification value is within the preset heart sound stability characteristic threshold range; If the heart sound stability classification value is within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a stable heart sound signal; If the heart sound stability classification value is not within the preset heart sound stability characteristic threshold range, the heart sound stability difference characteristic result is recorded as a non-stationary heart sound signal; The heart sound stationary difference characteristic result includes a stationary heart sound signal and a non-stationary heart sound signal.

4. The heart sound data processing method based on the time-frequency feature map according to claim 3, characterized in that: The specific steps of performing the heart sound optimization process based on the heart sound stationary difference characteristic result are as follows: If the heart sound stability difference characteristic result is a stable heart sound signal, performing stable heart sound optimization processing on the stable heart sound signal; The said steady heart sound optimization processing means optimizing the steady heart sound signal input based on the steady difference characteristic result of the heart sound to improve the accuracy of the heart sound signal in the process of murmur recognition; If the heart sound stability difference characteristic result is a non-stationary heart sound signal, performing non-stationary heart sound optimization processing on the non-stationary heart sound signal; The non-stationary heart sound optimization processing means optimizing the non-stationary heart sound signal input based on the heart sound stationary difference characteristic result to improve the accuracy of the heart sound signal in the process of murmur recognition; The heart sound optimization processing includes steady heart sound optimization processing and non-steady heart sound optimization processing.

5. The heart sound data processing method based on the time-frequency feature map according to claim 4, characterized in that: The specific process of performing the steady heart sound optimization processing on the steady heart sound signal is as follows: The steady heart sound signal is divided into short-term steady heart sound signals by the length of the steady time window, and a matching detection is performed based on the overlap between the short-term steady heart sound signal and a reference known murmur signal to obtain an overlap matching detection value, which is used to quantify the matching degree of the overlap between the short-term steady heart sound signal and the reference known murmur signal; The stable time window length represents the result of mapping the real-time heart sound stable division value and the deviation degree of the heart sound stable reference threshold into a stable division length mapping set in a preset database, wherein the stable division length mapping set represents the mapping relationship between the heart sound stable division value, the deviation degree of the heart sound stable reference threshold and the stable time window length; If the degree of deviation between the overlap match detection value and the reference overlap match threshold is within a preset allowable deviation range obtained from a preset database, then obtaining a murmur level recognition input priority for the short-time stationary heart sound signals based on the input confidence scores of the short-time stationary heart sound signals in descending order, and inputting a preset number of short-time stationary heart sound signals into the constructed murmur level recognition model according to the murmur level recognition input priority to obtain a murmur level recognition result; The input confidence score represents the result of mapping the deviation degree between the real-time overlap match detection value and the reference overlap match threshold into an input confidence mapping set in a preset database, wherein the input confidence mapping set represents the mapping relationship between the deviation degree between the overlap match detection value and the reference overlap match threshold and the input confidence score; If the degree of deviation between the overlap match detection value and the reference overlap match threshold is not within a preset allowable deviation range obtained from a preset database, principal component analysis is performed based on the feature dimensionality reduction strength of the short-time stationary heart sound signal, and a preset number of short-time stationary heart sound signals after principal component analysis and dimensionality reduction are input into the established murmur level recognition model to obtain a murmur level recognition result; The feature dimensionality reduction strength represents the result of mapping the degree of deviation between the real-time overlapping matching detection value and the reference overlapping matching threshold into the dimensionality reduction strength mapping set in the preset database. The dimensionality reduction strength mapping set represents the mapping relationship between the degree of deviation between the overlapping matching detection value and the reference overlapping matching threshold and the feature dimensionality reduction strength.

6. The heart sound data processing method based on the time-frequency characteristic graph according to claim 5, characterized in that: The matching detection is performed based on the overlap between the short-term steady heart sound signal and the reference known noise signal to obtain the overlap matching detection value. The specific steps are as follows: Acquiring overlap detection quantitative data between the short-term steady heart sound signal and the reference known murmur signal, wherein the overlap detection quantitative data includes a detection mutual correlation coefficient, a detection similarity coefficient, and a spectrum overlap coefficient; The detection cross-correlation coefficient is used to describe the similarity between the short-term stationary heart sound signal and the reference known murmur signal in the time domain, the detection similarity coefficient is used to describe the structural similarity between the short-term stationary heart sound signal and the reference known murmur signal, and the spectrum overlap coefficient is used to describe the similarity between the short-term stationary heart sound signal and the reference known murmur signal in the frequency domain; The overlap detection quantized data and the corresponding detection compensation amount are weighted and coupled to obtain the overlap matching detection value. The detection compensation amount includes the cross-correlation detection compensation amount, the similarity detection compensation amount and the spectrum overlap compensation amount; The overlap matching detection value represents the quantified data of the overlap matching detection of the short-term steady heart sound signal and the reference known noise signal.

7. The heart sound data processing method based on the time-frequency characteristic graph according to claim 4, characterized in that: The specific process of performing non-stationary heart sound optimization processing on the non-stationary heart sound signal is as follows: Determine whether the deviation between the heart sound stability classification value and the heart sound stability reference threshold is greater than a stability deviation setting value obtained from a preset database; If the degree of deviation between the heart sound stable division value and the heart sound stable reference threshold is greater than the stable deviation setting value obtained from a preset database, the degree of deviation between the heart sound stable division value and the stable deviation setting value is input into a heart sound sampling mapping set in the preset database for mapping to obtain a heart sound signal sampling frequency, and the heart sound signal is collected at the heart sound signal sampling frequency; the heart sound sampling mapping set represents a mapping relationship between the heart sound stable division value and the degree of deviation from the stable deviation setting value and the heart sound signal sampling frequency; If the deviation between the heart sound stability division value and the heart sound stability reference threshold is not greater than the stability deviation setting value obtained from the preset database, the non-stationary heart sound signal is divided into short-term non-stationary heart sound signals according to the non-stationary time window length; The non-stationary time window length represents the result of mapping the deviation degree between the real-time heart sound stationary division value and the heart sound stationary reference threshold into a non-stationary division length mapping set in a preset database, wherein the non-stationary division length mapping set represents the mapping relationship between the heart sound stationary division value, the deviation degree between the heart sound stationary reference threshold and the non-stationary time window length; Obtaining a signal energy value of the short-term non-stationary heart sound signal, and determining whether the signal energy value of the short-term non-stationary heart sound signal is greater than a reference signal energy threshold obtained from a preset database; If the signal energy value of the short-term non-stationary heart sound signal is greater than the reference signal energy threshold, the short-term non-stationary heart sound signal is marked as a valid heart sound signal; otherwise, the short-term non-stationary heart sound signal is marked as a waiting heart sound signal; Determine whether the short-term non-stationary heart sound signals in adjacent non-stationary time windows of the valid heart sound signal are all valid heart sound signals, where the length of the adjacent non-stationary time window includes the left adjacent non-stationary time window and the right adjacent non-stationary time window of the valid heart sound signal; If the short-term non-stationary heart sound signals within the non-stationary time window length adjacent to the valid heart sound signal are all valid heart sound signals, then the valid heart sound signal is marked as a valid priority heart sound signal; otherwise, no marking is performed; The effective priority heart sound signal is input into the constructed murmur level recognition model to obtain the murmur level recognition result.

8. The heart sound data processing method based on the time-frequency feature map according to claim 7, characterized in that: The specific process of verifying the noise level recognition result is as follows: If the heart sound stability difference characteristic result is a stable heart sound signal, a preset multiple of short-term stable heart sound signals are input into the established murmur level recognition model to obtain a verified murmur level recognition result; If the heart sound stability difference characteristic result is a non-stationary heart sound signal, a preset multiple number of valid heart sound signals are input into the established murmur level recognition model to obtain a verified murmur level recognition result; Determining whether the degree of conformity between the noise level recognition result and the verification noise level recognition result is greater than a preset conformity threshold obtained from a preset database; If the degree of conformity between the noise level recognition result and the verification noise level recognition result is greater than a preset conformity threshold obtained from a preset database, the verification is successful; otherwise, the noise level recognition result is updated.

9. The heart sound data processing method based on the time-frequency feature map according to claim 8, characterized in that: The specific steps of updating the noise level recognition result are: Verify the noise level recognition result. If the verification is still unsuccessful, verify the output results in turn. If the verification is successful, update the noise level recognition result to the output result for verification. If the verification is still unsuccessful when the number of input heart sound signals reaches the preset maximum number, the murmur level recognition result is updated to recognition abnormality.

10. A heart sound data processing system based on a time-frequency characteristic graph, characterized in that: include: Heart sound stable division module, optimized recognition processing module and murmur recognition and verification module; The heart sound steady division module is used to obtain the heart sound signal within a preset time interval, perform characteristic steady difference division on the time-frequency feature graph obtained based on the heart sound signal, and obtain the heart sound steady difference characteristic result; The optimization recognition processing module is used to perform heart sound optimization processing based on the heart sound stationary difference characteristic result to obtain a murmur level recognition result. The heart sound optimization processing means optimizing the heart sound signal input in the murmur recognition process based on the heart sound stationary difference characteristic result to improve the accuracy of murmur recognition of the heart sound signal; The noise recognition and verification module is used to verify the noise level recognition result. If the verification is successful, the noise level recognition result is fed back; otherwise, the noise level recognition result is updated.

Images