Respiratory Sound Recognition Method, Device, Equipment and Medium

Abstract

The present invention discloses a method, device, equipment and medium for respiratory sound recognition, including: obtaining a respiratory sound to be detected, preprocessing the respiratory sound to be detected to obtain a to-be-processed audio signal; performing wireless conversion on the to-be-processed audio signal and obtaining a wireless audio signal corresponding to the respiratory sound to be detected through wireless transmission; performing envelope transformation processing on the wireless audio signal to obtain an envelope feature result; performing anomaly detection on the envelope feature result through a respiratory sound anomaly detection model to obtain an anomaly detection result, where the anomaly detection result characterizes whether the respiratory sound to be detected is a normal respiratory sound; when the anomaly detection result characterizes a normal respiratory sound, performing respiratory sound classification and recognition on the envelope feature result through a respiratory sound recognition model to obtain a respiratory sound category corresponding to the respiratory sound to be detected. In this way, real-time detection of respiratory sounds is achieved, the accuracy of anomaly detection and judgment of respiratory sounds is improved, and the accuracy of respiratory sound recognition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sound recognition, and in particular to a method, device, equipment and medium for respiratory sound recognition. Background Art

[0002] Respiratory sound is a sound generated through structures such as airways and alveoli during the gas circulation process. It is an extremely important biological signal that can reflect important indicators of human-related information. For example, through the connection between respiratory sound and bispectral index, the bispectral index can be reflected. The common method is to listen to the respiratory sound through a stethoscope device and judge the abnormality of the respiratory sound based on experience. Although the existing technical solutions can judge the abnormality according to the neural network model, because the existing stethoscope device consists of a sound pickup part, a conduction part and a listening part, it needs to hold the sound pickup part for collection, resulting in limitations on the distance of the stethoscope device and being easily affected by other noises during the collection process, so that the human respiratory sound cannot be detected in real time and the accurate recognition of the respiratory sound category cannot be achieved. Summary of the Invention

[0003] The present invention provides a method, device, equipment and medium for respiratory sound recognition, which realizes the real-time detection of respiratory sound, breaks through the limitation of the conduction part on the detection of respiratory sound through wireless transmission, improves the accuracy of abnormal detection of respiratory sound, and improves the accuracy of recognition of respiratory sound categories.

[0004] A method for respiratory sound recognition includes:

[0005] Obtain the respiratory sound to be detected, preprocess the respiratory sound to be detected to obtain a to-be-processed audio signal;

[0006] Perform wireless conversion on the to-be-processed audio signal, and obtain a wireless audio signal corresponding to the respiratory sound to be detected through wireless transmission;

[0007] Perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result;

[0008] Perform abnormal detection on the envelope feature result through a respiratory sound abnormal detection model to obtain an abnormal detection result; the abnormal detection result characterizes whether the respiratory sound to be detected is a normal respiratory sound;

[0009] When the abnormal detection result characterizes a normal respiratory sound, perform respiratory sound classification recognition on the envelope feature result through a respiratory sound recognition model to obtain a respiratory sound category corresponding to the respiratory sound to be detected.

[0010] A respiratory sound recognition device includes:

[0011] An acquisition module, configured to acquire a breath sound to be detected, preprocess the breath sound to be detected, and obtain a to-be-processed audio signal;

[0012] A conversion module, configured to perform wireless conversion on the to-be-processed audio signal, and wirelessly transmit the signal to obtain a wireless audio signal corresponding to the breath sound to be detected;

[0013] A transformation module, configured to perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result;

[0014] A detection module, configured to perform anomaly detection on the envelope feature result through a breath sound anomaly detection model to obtain an anomaly detection result, where the anomaly detection result characterizes whether the breath sound to be detected is a normal breath sound;

[0015] An identification module, configured to, when the anomaly detection result characterizes a normal breath sound, perform breath sound classification and identification on the envelope feature result through a breath sound identification model to obtain a breath sound category corresponding to the breath sound to be detected.

[0016] A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where when the processor executes the computer program, the above-mentioned breath sound identification method is implemented.

[0017] A computer-readable storage medium storing a computer program, where when the computer program is executed by a processor, the above-mentioned breath sound identification method is implemented.

[0018] The breath sound identification method, device, equipment, and medium provided by the present invention obtain a breath sound to be detected, preprocess the breath sound to be detected to obtain a to-be-processed audio signal; perform wireless conversion on the to-be-processed audio signal and wirelessly transmit the signal to obtain a wireless audio signal corresponding to the breath sound to be detected; perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result; perform anomaly detection on the envelope feature result through a breath sound anomaly detection model to obtain an anomaly detection result, where the anomaly detection result characterizes whether the breath sound to be detected is a normal breath sound; when the anomaly detection result characterizes a normal breath sound, perform breath sound classification and identification on the envelope feature result through a breath sound identification model to obtain a breath sound category corresponding to the breath sound to be detected; in this way, real-time detection of breath sounds is achieved, the limitation of the conduction part on detecting breath sounds is broken through by wireless transmission, interference from other sounds is reduced, thereby improving the accuracy of breath sound anomaly detection and the accuracy of breath sound category identification, and making the operation of users more convenient. Description of the Drawings

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the accompanying drawings required for the description of the embodiments of the present invention. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 is a flowchart of a breath sound recognition method in an embodiment of the present invention;

[0021] Figure 2 is a flowchart of step S10 of the breath sound recognition method in an embodiment of the present invention;

[0022] Figure 3 is a flowchart of step S30 of the breath sound recognition method in an embodiment of the present invention;

[0023] Figure 4 is a flowchart of step S40 of the breath sound recognition method in an embodiment of the present invention;

[0024] Figure 5 is a flowchart of step S50 of the breath sound recognition method in an embodiment of the present invention;

[0025] Figure 6 is a flowchart of step S50 of the breath sound recognition method in another embodiment of the present invention;

[0026] Figure 7 is a principle block diagram of a breath sound recognition device in an embodiment of the present invention;

[0027] Figure 8 is a schematic diagram of a computer device in an embodiment of the present invention. Detailed implementation manners

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0029] The present invention provides a breath sound recognition method. In one embodiment, as Figure 1 shown, its technical solutions mainly include the following steps S10 - S50:

[0030] S10, obtain the breath sound to be detected, preprocess the breath sound to be detected, and obtain the audio signal to be processed.

[0031] Understandably, the breathing sound during the human breathing process is collected by a breathing sound collection device. The breathing sound collection device can be in the shape of a disc, a square, a rectangle, etc. The breathing sound collection device includes a tympanic membrane and a resonance cavity made of TPU flexible material. The breathing sound collection device is fixed to the human neck by pasting. The TPU tympanic membrane collects the breathing sound of the human body and amplifies it in the resonance cavity to obtain the breathing sound to be detected. The breathing sound to be detected is the breathing sound of the human body collected by the breathing sound collection device. The process of preprocessing the breathing sound to be detected is to obtain the amplified breathing sound through a pickup, convert the breathing sound into an analog signal, and after amplifying and filtering the analog signal through an audio chip, a digital signal is obtained, and this digital signal is recorded as the audio signal to be processed. The pickup preferably selects a MEMS pickup. The audio signal to be processed is the digital signal preprocessed by the pickup and the audio chip.

[0032] In this way, the breathing sound of the human neck is collected by the tympanic membrane and resonance cavity made of TPU material, the breathing sound is obtained through a pickup and converted into an analog signal, and the audio chip converts the analog breathing sound signal into a digital signal, that is, the audio signal to be processed. In this way, the collected breathing sound to be detected is made clearer, the influence caused by the breathing sound of other parts is reduced, and the real-time detection of the human breathing sound is realized.

[0033] In one embodiment, as Figure 2 shown, in step S10, that is, preprocessing the breathing sound to be detected to obtain the audio signal to be processed, includes:

[0034] S101, performing analog conversion on the breathing sound to be detected to obtain a breathing sound analog signal.

[0035] Understandably, the collected breathing sound to be detected is subjected to analog conversion through a pickup, and the breathing sound to be detected is converted into an analog signal, that is, a breathing sound analog signal. The analog conversion is to convert the breathing sound to be detected into an analog signal.

[0036] S102, performing audio conversion processing on the breathing sound analog signal to obtain the audio signal to be processed.

[0037] Understandably, the breathing sound analog signal is subjected to audio conversion processing through an audio chip. The audio chip converts the audio signal into a digital signal. The audio chip converts the breathing sound analog signal through amplification and filtering processing to obtain a digital signal, and this digital signal is recorded as the audio signal to be processed. The audio conversion is to convert an analog signal into an audio signal.

[0038] The present invention realizes converting the to-be-detected breath sound into a breath sound analog signal through a pickup, and converting the breath sound analog signal into a digital signal by an audio chip, that is, the audio signal to be processed. In this way, the real-time acquisition of the breath sound and the conversion of the breath sound are realized, making the obtained breath sound clearer and reducing the influence caused by other sounds.

[0039] S20, wirelessly convert the to-be-processed audio signal, and obtain a wireless audio signal corresponding to the to-be-detected breath sound through wireless transmission.

[0040] Understandably, the to-be-processed audio signal is wirelessly converted by a portable acquisition device, and is transmitted by a processor through a wireless transmission method to obtain a wireless audio signal corresponding to the to-be-detected breath sound. The portable acquisition device includes a breath sound acquisition device, a pickup, and an audio chip. The portable acquisition device sends the wireless audio signal to a smart terminal or a mobile terminal. The wireless conversion is a way of converting data using wireless technology. The wireless technology can be technologies such as WIFI wireless technology, infrared wireless technology, and Bluetooth wireless technology. The processor can be processors such as an MCU processor, a DSP processor, and an FPGA processor. The wireless transmission is the transmission of data through wireless technology. The wireless audio signal is a digital signal obtained by preprocessing the to-be-detected breath sound and then performing wireless conversion. The audio signal is a carrier of the frequency and amplitude change information of a sound wave with a vibration pattern.

[0041] S30, perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result.

[0042] Understandably, after a smart terminal or a mobile terminal receives the wireless audio signal corresponding to the to-be-detected breath sound through wireless transmission, the wireless audio signal is subjected to envelope transformation processing. The envelope transformation is to extract the envelope of the wireless audio signal to obtain an envelope signal, and extract the features of the envelope signal to obtain an envelope feature result. The envelope signal is an array after extracting the signal envelope. The envelope feature result is the result after extracting envelope parameters from the envelope signal to construct a feature vector, that is, the envelope feature result is a set of envelope-related features selected in advance. The features of the envelope signal are features such as the maximum value, mean value, standard deviation, variance, kurtosis, skewness, and energy of the preset envelope.

[0043] In an embodiment, as Figure 3 shown, in the step S30, that is, performing envelope transformation processing on the wireless audio signal to obtain an envelope feature result, includes:

[0044] S301, Based on a preset frequency band, perform modal decomposition and noise removal on the wireless audio signal to obtain an audio modal signal.

[0045] Understandably, decompose and denoise the wireless audio signal according to a preset frequency band, denoise the wireless audio signal through a band-pass filter to remove high-frequency noise and low-frequency noise, obtain a wireless audio signal in the range of 100 - 1200 Hz, and then perform modal decomposition on the wireless audio signal in the range of 100 - 1200 Hz to obtain the intrinsic mode functions in the range of 100 - 1200 Hz, that is, the audio modal signal. The high-frequency noise is the sound greater than 1200 Hz, the low-frequency noise is the sound less than 100 Hz. The band-pass filter is a component that allows waves in a specific frequency band to pass through while blocking other frequency bands. The modal decomposition is to decompose the signal according to the time-scale characteristics of the data itself. The intrinsic mode function has a meaningful instantaneous frequency. Then the function must be symmetric, have a local mean of zero, and have the same number of zero-crossing points and extreme points.

[0046] S302, Perform envelope signal conversion processing on the audio modal signal to obtain an envelope signal.

[0047] Understandably, after obtaining the audio modal signal, perform envelope conversion processing on the audio modal signal to obtain an envelope signal. The envelope conversion is the signal obtained by extracting the envelope of the audio modal signal. The envelope is an equi-amplitude oscillating pulse signal. After modulation, the amplitude of each oscillation will change. Connect the highest point and the lowest point of each oscillation signal with a curve respectively. The functional expression of the curve is the envelope of the pulse signal.

[0048] S303, Extract envelope parameters from the envelope signal to obtain the envelope feature result.

[0049] Understandably, according to preset characteristic parameters, extract the envelope parameters in the envelope signal to obtain the envelope feature result. The envelope parameters are the parameters of the envelope signal, such as parameters like the maximum value, mean value, standard deviation, variance, kurtosis, skewness, energy, etc. of the envelope. The envelope feature result is the result after extracting the envelope parameters of the envelope signal to construct a feature vector.

[0050] The present invention realizes noise processing and signal decomposition of the wireless audio signal through modal decomposition and a band-pass filter, performs envelope conversion processing on the decomposed signal to obtain an envelope signal, extracts parameters from the envelope signal to construct a feature vector, and obtains the envelope feature result. In this way, the removal of signal noise and the decomposition of the signal are realized, improving the accuracy of the abnormal judgment of the to-be-detected breath sound.

[0051] In one embodiment, step S302, that is, performing envelope signal conversion processing on the audio modal signal to obtain an envelope signal, includes:

[0052] Applying the Hilbert algorithm to perform a Hilbert transform on the audio modal signal to obtain the envelope signal.

[0053] Understandably, after obtaining the audio modal signal, by applying the Hilbert algorithm to perform a Hilbert transform on the audio modal signal to obtain an envelope signal, the Hilbert transform is to convolve the signal s(t) with 1 / (πt) to obtain s'(t). The Hilbert transform result s'(t) can be interpreted as the output of a linear time invariant system whose input is s(t), and the impulse response of this system is 1 / (πt). The envelope signal is the array of the audio modal signal after the Hilbert transform. The Hilbert transform process is to construct an analytic signal, aiming to transform a real signal into a complex signal. Let the original signal be the real part and the signal after the Hilbert transform be the imaginary part. Since the signal has both amplitude information and phase information, the analytic signal can be set and substituted. Taking the absolute value of the analytic signal is the envelope signal.

[0054] The present invention realizes obtaining the envelope signal by applying the Hilbert algorithm to perform a Hilbert transform on the audio modal signal. In this way, complex operations are avoided, the extraction of signal envelope parameters is facilitated, and the analysis and transformation of non-linear and non-stationary signals are realized.

[0055] In another embodiment, step S302, that is, performing envelope signal conversion processing on the audio modal signal to obtain an envelope signal, includes:

[0056] Applying the wavelet packet algorithm to perform a Morlet transform on the audio modal signal to obtain the envelope signal.

[0057] Understandably, after obtaining the audio modal signal, the Morlet transform is performed on the audio modal signal by applying the wavelet packet algorithm to obtain an envelope signal. The wavelet transform selects a wavelet basis function (corresponding to the central frequency w0), obtains a series of central frequencies (w0 / a) through scale transformation, and then obtains a series of basis functions in different intervals through time shift. Then, the product of each basis function and the original signal (in the corresponding interval) is integrated, and the frequency corresponding to the generated extreme value is the frequency of the original signal in this interval. That is, the wavelet transform is a local transform in the time and frequency domains, so it can effectively extract information from the signal and perform multi-scale refinement analysis on the function or signal through operations such as stretching and translation. The process of the wavelet transform is to first use the audio modal signal as the input signal, decompose it into a high-frequency part and a low-frequency part through a set of orthogonal wavelet bases, and then use the obtained low-frequency part as the input signal and perform wavelet decomposition again to obtain the next-level high-frequency part and low-frequency part. As the number of levels of wavelet decomposition increases, its resolution in the frequency domain becomes higher.

[0058] The present invention realizes the Morlet transform of the audio modal signal by applying the wavelet packet algorithm to obtain the envelope signal. In this way, the extraction of the signal envelope and the acquisition of the envelope signal are realized, which facilitates the subsequent processing of the signal, reduces the correlation between different extracted features, and realizes the fast transformation of the audio modal signal.

[0059] S40. Perform anomaly detection on the envelope feature result through a breath sound anomaly detection model to obtain an anomaly detection result, and the anomaly detection result characterizes whether the breath sound to be detected is a normal breath sound.

[0060] Understandably, the envelope feature result is input into the breath sound anomaly detection model for breath sound anomaly detection. The breath sound anomaly detection model outputs an anomaly detection result, and the anomaly detection result represents whether the breath sound to be detected is a normal breath sound. The breath sound anomaly detection model is a neural network model for breath sound anomaly detection. The process of the anomaly detection is to extract breath sound anomaly features from the envelope feature result and determine the anomaly detection result according to the extracted breath sound anomaly features. The network structure of the breath sound anomaly detection model is set according to requirements. For example, the network structure of the breath sound anomaly detection model can be the network structure of a backpropagation neural network (BP). The network structure of the breath sound anomaly detection model can also be the network structure of an RNN, a CNN, or a WaveUNet. The network structure of the WaveUNet is a multi-scale neural network for end-to-end audio source separation. The anomaly detection result is the result output by the breath sound anomaly detection model after detecting the breath sound to be detected.

[0061] Among them, the abnormal detection result output by the breath sound abnormal detection model includes normal breath sound and abnormal breath sound. When the abnormal detection result is an abnormal breath sound, the set abnormal alarm is triggered to remind the detection personnel. When the abnormal detection result is a normal breath sound, go to step S50, and input the envelope feature result into the breath sound recognition model for recognition and classification.

[0062] In one embodiment, as Figure 4 shown, in step S40, that is, the envelope feature result is subjected to abnormal detection by the breath sound abnormal detection model to obtain an abnormal detection result, and the abnormal detection result characterizes whether the breath sound to be detected is a normal breath sound, including:

[0063] S401, extract the abnormal breath sound features of the envelope feature result through the breath sound abnormal detection model to obtain a feature array.

[0064] Understandably, after obtaining the envelope feature result, the breath sound abnormal detection model extracts the abnormal breath sound features of the envelope feature result and organizes them into a feature array. The abnormal breath sound features are preset features such as the maximum value, mean value, standard deviation, variance, kurtosis, skewness, and energy of the envelope, and the feature array is a sequence combination of the abnormal breath sound features.

[0065] S402, perform abnormal detection of the breath sound on the feature array through the breath sound abnormal detection model to obtain the abnormal detection result.

[0066] Understandably, the feature array is subjected to abnormal detection in the breath sound abnormal detection model, and the breath sound abnormal detection model outputs the abnormal detection result of the feature array. The abnormal detection result includes abnormal breath sound and normal breath sound. The network structure of the breath sound abnormal detection model can be the same as or different from the network structure of the breath sound recognition model.

[0067] The present invention realizes extracting the abnormal breath sound features of the envelope feature result through the breath sound abnormal detection model to obtain a feature array, and performing abnormal detection of the breath sound on the feature array through the breath sound abnormal detection model to obtain the abnormal detection result. In this way, the judgment of the breath sound abnormal detection is realized, and the accuracy of the breath sound abnormal detection judgment is improved.

[0068] S50, when the abnormal detection result is characterized as a normal breath sound, perform breath sound classification and recognition on the envelope feature result through the breath sound recognition model to obtain the breath sound category corresponding to the breath sound to be detected.

[0069] Understandably, when the abnormal detection result output by the respiratory sound abnormality detection model is normal respiratory sound, the envelope feature result is input into the respiratory sound recognition model for respiratory sound classification and recognition to obtain the respiratory sound category corresponding to the respiratory sound to be detected. The respiratory sound category includes four situations: smooth, gentle, slow, and sudden change.

[0070] Among them, the respiratory sound recognition model is a neural network model constructed by a backpropagation neural network. Multi-period respiratory sound data corresponding to the bispectral index of the electroencephalogram and multi-period respiratory sound data of other abnormal respiratory sounds are obtained. Part of the respiratory sound data is selected as training data to construct a respiratory sound recognition model. The part of the respiratory sound data can be 30%, 50%, or 70% of the total data collected historically. The part of the respiratory sound data is used as the training sample of the respiratory sound recognition model for deep learning training. The part of the respiratory sound data is input into the initial respiratory sound recognition model based on the backpropagation neural network for learning, and finally the activation function is output, successfully constructing the respiratory sound recognition model. The remaining respiratory sound data or all the respiratory sound data after selection is used as test data to test the constructed respiratory sound recognition model. When the test rate exceeds the set recognition rate, the test is completed and the respiratory sound recognition model is output. When the test rate does not exceed the set recognition rate, the remaining respiratory sound data or all the respiratory sound data after selection is re-input for recognition, and the weight parameters in the respiratory sound recognition model are iteratively updated until the test recognition rate exceeds the set recognition rate, then the test is completed and the respiratory sound recognition model is output. The set recognition rate can be set according to actual needs, such as a recognition rate of 90% or a recognition rate of 95%.

[0071] The respiratory sound recognition method, device, equipment, and medium provided by the present invention obtain the respiratory sound to be detected, preprocess the respiratory sound to be detected to obtain a to-be-processed audio signal; perform wireless conversion on the to-be-processed audio signal and obtain a wireless audio signal corresponding to the respiratory sound to be detected through wireless transmission; perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result; perform abnormal detection on the envelope feature result through a respiratory sound abnormality detection model to obtain an abnormal detection result, and the abnormal detection result characterizes whether the respiratory sound to be detected is a normal respiratory sound; when the abnormal detection result characterizes a normal respiratory sound, perform respiratory sound classification and recognition on the envelope feature result through a respiratory sound recognition model to obtain the respiratory sound category corresponding to the respiratory sound to be detected; in this way, real-time detection of respiratory sound is achieved, the limitation of the conduction part on detecting respiratory sound is broken through through wireless transmission, the interference of other sounds is reduced, thereby improving the accuracy of respiratory sound abnormality detection and the accuracy of respiratory sound category recognition, and making the operation of users more convenient.

[0072] In one embodiment, as Figure 5 shown, in step 50, that is, when the abnormal detection result represents normal breath sounds, the envelope feature result is classified and recognized for breath sounds through a breath sound recognition model to obtain a breath sound category corresponding to the breath sound to be detected, including:

[0073] S501, extracting breath sound level features from the envelope feature result through the breath sound recognition model, and performing activation processing on the extracted breath sound level features to obtain a feature vector map.

[0074] Understandably, after inputting the envelope feature result into the breath sound recognition model, the breath sound recognition model extracts the breath sound level features corresponding to the envelope feature result, and performs activation processing on the breath sound level features in the breath sound recognition model, that is, processes the breath sound level features with an activation function to obtain a feature vector map corresponding to the envelope feature signal. The feature vector map is an image after extracting feature vectors from the envelope feature signal, and the breath sound level features are features corresponding to different breath sound categories.

[0075] S502, performing breath sound classification processing on the feature vector map through the breath sound recognition model to obtain the breath sound category.

[0076] Understandably, the breath sound recognition model performs breath sound classification processing on the feature vector map. The breath sound classification processing is to perform probability classification of the full connection layer on the feature vector map, predict the prediction probabilities of the feature vector map corresponding to each breath sound category, and determine the breath sound category corresponding to the maximum prediction probability as the breath sound category corresponding to the breath sound to be detected.

[0077] The present invention realizes extracting breath sound level features from the envelope feature result, performing activation processing, obtaining a feature vector map corresponding to the envelope feature signal, performing breath sound classification processing on the feature vector map, and obtaining the breath sound category. In this way, the recognition of the breath sound category to be detected is realized, the accuracy of the recognition of the breath sound category to be detected is improved, and the operation of the user is further made more convenient.

[0078] In one embodiment, as Figure 6 shown, after step 50, that is, when the abnormal detection result represents normal breath sounds, after classifying and recognizing the envelope feature result for breath sounds through a breath sound recognition model to obtain a breath sound category corresponding to the breath sound to be detected, it includes:

[0079] S503, determining an electroencephalogram bispectral index corresponding to the breath sound to be detected according to the breath sound category.

[0080] Understandably, after obtaining the respiratory sound category, look up the bispectral index corresponding to the respiratory sound category in the respiratory sound category table. The bispectral index corresponding to smooth is 100, the bispectral index corresponding to gentle is 75, the bispectral index corresponding to sluggish is 50, and the bispectral index corresponding to mutation is 25. The bispectral index refers to measuring the linear components (frequency and power) of the electroencephalogram, analyzing the non-linear relationship (phase and harmonic) between component waves, selecting various electroencephalogram signals representing different sedation levels, performing standardization and digitization processing, and finally converting them into a simple quantitative index.

[0081] S504, play the bispectral index corresponding to the to-be-detected respiratory sound and the wireless audio signal.

[0082] Understandably, after obtaining the bispectral index and the wireless audio signal, they can be played through a mobile terminal or a smart terminal for reminder. The smart terminal can be terminal devices such as a smart phone, a Bluetooth speaker, and a computer. Among them, the range of the bispectral index corresponding to smooth displayed on the smart terminal or the mobile terminal is 100 - 85, the range of the bispectral index corresponding to gentle is 85 - 65, the range of the bispectral index corresponding to sluggish is 65 - 40, and the range of the bispectral index corresponding to mutation is 40 - 0.

[0083] The present invention realizes determining the corresponding bispectral index through the respiratory sound category, and playing the bispectral index and the wireless audio signal through a smart terminal or a mobile terminal. In this way, it realizes the category recognition of the to-be-detected respiratory sound, realizes the playing of the bispectral index and the wireless audio signal corresponding to the to-be-detected respiratory sound, reminds the respiratory state of the currently detected person, facilitates the preparation for subsequent work, and improves the user operation experience.

[0084] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0085] In an embodiment, a respiratory sound recognition device is provided, and this respiratory sound recognition device corresponds one-to-one with the respiratory sound recognition method in the above embodiment. As Figure 7 shown, this respiratory sound recognition device includes an acquisition module 11, a conversion module 12, a transformation module 13, a detection module 14, and a recognition module 15. The detailed description of each functional module is as follows:

[0086] The acquisition module 11 is used to acquire the to-be-detected respiratory sound, preprocess the to-be-detected respiratory sound, and obtain a to-be-processed audio signal;

[0087] A conversion module 12, configured to perform wireless conversion on the to-be-processed audio signal, and obtain a wireless audio signal corresponding to the to-be-detected breath sound through wireless transmission;

[0088] A transformation module 13, configured to perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result;

[0089] A detection module 14, configured to perform anomaly detection on the envelope feature result through a breath sound anomaly detection model to obtain an anomaly detection result, where the anomaly detection result characterizes whether the to-be-detected breath sound is a normal breath sound;

[0090] An identification module 15, configured to, when the anomaly detection result characterizes a normal breath sound, perform breath sound classification and identification on the envelope feature result through a breath sound identification model to obtain a breath sound category corresponding to the to-be-detected breath sound.

[0091] In this way, the present invention realizes real-time detection of the to-be-detected breath sound, realizes real-time detection of the human breath sound, breaks through the limitation of the conduction part on detecting the breath sound through wireless transmission, further improves the accuracy of breath sound anomaly detection and judgment, and improves the accuracy of breath sound category identification, making the operation of the user more convenient.

[0092] In one embodiment, the acquisition module includes:

[0093] An analog conversion unit, configured to perform analog conversion on the to-be-detected breath sound to obtain a breath sound analog signal;

[0094] An audio conversion unit, configured to perform audio conversion processing on the breath sound analog signal to obtain the to-be-processed audio signal.

[0095] In one embodiment, the transformation module includes:

[0096] A decomposition and removal unit, configured to perform modal decomposition and noise removal on the wireless audio signal based on a preset frequency band to obtain an audio modal signal;

[0097] An envelope conversion unit, configured to perform envelope signal conversion processing on the audio modal signal to obtain an envelope signal;

[0098] A parameter extraction unit, configured to perform envelope parameter extraction on the envelope signal to obtain the envelope feature result.

[0099] In one embodiment, the transformation module further includes:

[0100] A first transformation unit, configured to perform Hilbert transform on the audio modal signal by using the Hilbert algorithm to obtain the envelope signal;

[0101] A second transformation unit, configured to perform Morlet transformation on the audio modal signal by using a wavelet packet algorithm to obtain the envelope signal.

[0102] In one embodiment, the detection module includes:

[0103] A feature extraction unit, configured to extract abnormal breath sound features from the envelope feature result through the breath sound abnormality detection model to obtain a feature array;

[0104] A detection unit, configured to perform breath sound abnormality detection on the feature array through the breath sound abnormality detection model to obtain the abnormality detection result.

[0105] In one embodiment, the recognition module includes:

[0106] An activation unit, configured to extract breath sound level features from the envelope feature result through the breath sound recognition model and perform activation processing on the extracted breath sound level features to obtain a feature vector map;

[0107] A processing unit, configured to perform breath sound classification processing on the feature vector map through the breath sound recognition model to obtain the breath sound category.

[0108] In one embodiment, the recognition module further includes:

[0109] A determination unit, configured to determine an electroencephalogram bispectral index corresponding to the breath sound to be detected according to the breath sound category;

[0110] A playback unit, configured to play the electroencephalogram bispectral index and the wireless audio signal corresponding to the breath sound to be detected.

[0111] For the specific limitations of the breath sound recognition device, reference can be made to the limitations on the breath sound recognition method in the above text, which will not be elaborated here. Each module in the above breath sound recognition device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above modules.

[0112] In one embodiment, a computer device is provided. The computer device can be a client or a server, and its internal structure diagram can be as Figure 8As shown. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the readable storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements a breath sound recognition method.

[0113] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the breath sound recognition method in the above embodiment.

[0114] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the breath sound recognition method in the above embodiment.

[0115] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0117] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for respiratory sound recognition, characterized in that, Including: Obtain the respiratory sound to be detected, preprocess the respiratory sound to be detected to obtain a to-be-processed audio signal; wherein, the respiratory sound acquisition device is fixed on the neck of the human body by pasting, the TPU eardrum in the respiratory sound acquisition device collects the respiratory sound, and amplifies it in the resonance cavity; Perform wireless conversion on the to-be-processed audio signal, and obtain a wireless audio signal corresponding to the respiratory sound to be detected through wireless transmission; Perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result; Perform anomaly detection on the envelope feature result through a respiratory sound anomaly detection model to obtain an anomaly detection result, and the anomaly detection result characterizes whether the respiratory sound to be detected is a normal respiratory sound; When the anomaly detection result characterizes a normal respiratory sound, perform respiratory sound classification and recognition on the envelope feature result through a respiratory sound recognition model to obtain a respiratory sound category corresponding to the respiratory sound to be detected; After obtaining the respiratory sound category corresponding to the respiratory sound to be detected, including: Determine the bispectral index of electroencephalogram corresponding to the respiratory sound to be detected according to the respiratory sound category; Play the bispectral index of electroencephalogram and the wireless audio signal corresponding to the respiratory sound to be detected.

2. The respiratory sound recognition method according to claim 1, wherein The preprocessing the respiratory sound to be detected to obtain a to-be-processed audio signal includes: Perform analog conversion on the respiratory sound to be detected to obtain a respiratory sound analog signal; Perform audio conversion processing on the respiratory sound analog signal to obtain the to-be-processed audio signal.

3. The respiratory sound recognition method according to claim 1, characterized in that, The performing envelope transformation processing on the wireless audio signal to obtain an envelope feature result includes: Based on a preset frequency band, perform modal decomposition and noise removal on the wireless audio signal to obtain an audio modal signal; Perform envelope signal conversion processing on the audio modal signal to obtain an envelope signal; Extract envelope parameters from the envelope signal to obtain the envelope feature result.

4. The respiratory sound recognition method according to claim 3, wherein The performing envelope signal conversion processing on the audio modal signal to obtain an envelope signal includes: Apply the Hilbert algorithm to perform Hilbert transform on the audio modal signal to obtain the envelope signal; Or Apply the wavelet packet algorithm to perform Morlet transform on the audio modal signal to obtain the envelope signal.

5. The respiratory sound recognition method according to claim 1, wherein The performing anomaly detection on the envelope feature result through a respiratory sound anomaly detection model to obtain an anomaly detection result includes: Extract respiratory sound anomaly features from the envelope feature result through the respiratory sound anomaly detection model to obtain a feature array; Perform respiratory sound anomaly detection on the feature array through the respiratory sound anomaly detection model to obtain the anomaly detection result.

6. The respiratory sound recognition method according to claim 1, wherein The performing respiratory sound classification and recognition on the envelope feature result through a respiratory sound recognition model to obtain a respiratory sound category corresponding to the respiratory sound to be detected includes: Extract respiratory sound level features from the envelope feature result through the respiratory sound recognition model, and perform activation processing on the extracted respiratory sound level features to obtain a feature vector map; Perform respiratory sound classification processing on the feature vector map through the respiratory sound recognition model to obtain the respiratory sound category.

7. A respiratory sound recognition device, characterized in that, Including: An acquisition module, configured to acquire a breath sound to be detected, preprocess the breath sound to be detected, and obtain a to-be-processed audio signal; wherein, the breath sound acquisition device is fixed to the neck of a human body by pasting, and the TPU eardrum in the breath sound acquisition device collects the breath sound and amplifies it in the resonance cavity; A conversion module, configured to wirelessly convert the to-be-processed audio signal, and obtain a wireless audio signal corresponding to the breath sound to be detected through wireless transmission; A transformation module, configured to perform envelope transformation processing on the wireless audio signal to obtain an envelope feature result; A detection module, configured to perform anomaly detection on the envelope feature result through a breath sound anomaly detection model to obtain an anomaly detection result, and the anomaly detection result characterizes whether the breath sound to be detected is a normal breath sound; An identification module, configured to, when the anomaly detection result characterizes a normal breath sound, perform breath sound classification and identification on the envelope feature result through a breath sound identification model to obtain a breath sound category corresponding to the breath sound to be detected; The identification module further includes: A determination unit, configured to determine an electroencephalogram bispectral index corresponding to the breath sound to be detected according to the breath sound category; A playback unit, configured to play the electroencephalogram bispectral index and the wireless audio signal corresponding to the breath sound to be detected.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the breath sound identification method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the breath sound identification method according to any one of claims 1 to 6.

Images