Real-time cough sound counting method

Abstract

The invention discloses a real-time cough sound counting method, and relates to the technical field of audio recognition, and the method comprises the steps: determining a user side, collecting and processing the audio of the user side, and obtaining a plurality of target audios; extracting the plurality of target audios based on a neural network algorithm to obtain cough audios of the user side, and performing quantity statistics on the cough audios; and transmitting the quantity of the cough audios to a server as use data of the user. According to the system, the accuracy of cough sound recognition and extraction is improved, analysis data of the physical state of a user side person is better provided, the difficulty of collecting cough data of a patient suffering from chronic respiratory diseases is reduced by collecting the cough information of the patient through the system, and the accuracy and timeliness of collecting the cough data are guaranteed.

Description

Technical Field

[0001] The invention relates to the technical field of audio recognition, and in particular to a real-time cough sound counting method. Background Art

[0002] At present, with the air pollution, the number of patients with chronic respiratory diseases is gradually increasing. In order to accurately manage chronic respiratory diseases, cough has become an important biomarker for evaluating chronic respiratory diseases. The current cough assessment mainly includes visual analogue score (VAS), cough symptom score, quality of life assessment, cough frequency monitoring and cough sensitivity detection, etc. These methods have so far relied heavily on manual measurement to determine the cough frequency, which makes it difficult to reduce the difficulty of collecting cough data for patients with chronic respiratory diseases, and cannot guarantee the accuracy and timeliness of the collected cough data. Summary of the invention

[0003] The purpose of the present invention is to provide a real-time cough sound counting method to solve the shortcomings of the background technology.

[0004] In order to achieve the above object, the present invention provides the following technical solution: a real-time cough sound counting method, comprising the following steps:

[0005] Determine a user terminal, collect and process the audio of the user terminal to obtain multiple target audios;

[0006] Based on the neural network algorithm, multiple target audios are extracted to obtain the cough audios of the user end, and the number of cough audios is counted;

[0007] The number of cough audios is transmitted to the server as the user's usage data.

[0008] In a preferred embodiment, the step of determining the user terminal, collecting and processing the audio of the user terminal to obtain multiple target audios includes:

[0009] Determine user information, wherein the user information includes basic user information and basic user physical information, and obtain a user terminal according to the user information registration;

[0010] Formulate a user-side audio processing tree, wherein the audio processing tree includes an audio input channel, a classification channel, and a filtering point between the audio input channel and the classification channel;

[0011] The audio from the user end is collected and put into the audio processing tree in real time for processing to obtain multiple target audios.

[0012] In a preferred embodiment, the step of formulating the user-side audio processing tree includes:

[0013] Formulate an audio input channel, set multiple filter points in the audio input channel, the multiple filter points are interconnected, and different filter information is formulated corresponding to the multiple filter points;

[0014] Based on the filtering point, the classification channel is obtained by diverging through the audio input channel;

[0015] In the audio input channel, a spiral channel is set corresponding to the number of filter points and associated with the filter points, and the plurality of spiral channels are intercommunicated with each other at intervals;

[0016] The audio input channel, the classification channel, and the filtering points between the audio input channel and the classification channel are used as an audio processing tree.

[0017] In a preferred embodiment, the step of setting spiral channels in the audio input channel in accordance with the number of filter points and in association with the filter points, wherein the plurality of spiral channels are intermittently interconnected, comprises:

[0018] Set a corresponding number of spiral channels in the audio input channel corresponding to the number of filter points;

[0019] Establish a one-to-one correspondence between the filtering points and the spiral channels, and connect the spiral channels and the filtering points that have a one-to-one correspondence;

[0020] A plurality of interaction points of the same number are respectively selected in the plurality of spiral channels, and an interconnection relationship between the plurality of spiral channels is established based on the interaction points.

[0021] In a preferred embodiment, the step of collecting the audio of the user terminal and putting it into the audio processing tree in real time for processing to obtain multiple target audios includes:

[0022] The audio of the user end is collected in real time and input into the audio input channel, the timbre of the audio of the user end is recognized through the filter point to obtain multiple audio timbres, the multiple audio timbres are respectively assigned to the filter points, and the audio of the audio timbre corresponding to the filter point is separated in the spiral channel;

[0023] Continuously collect the user's audio in real time and separate the user's audio according to the audio timbre;

[0024] The spiral channel performs self-inspection to obtain a self-inspection result, and the self-inspection result is fed back to the filter point for confirmation. The audio is re-divided according to the confirmed self-inspection result, and the re-divided audio is input into the classification channel through the spiral channel to store multiple target audios.

[0025] In a preferred embodiment, the spiral channel performs self-test to obtain a self-test result, feeds the self-test result back to the filtering point for confirmation, and performs the step of re-dividing the audio according to the confirmed self-test result, including:

[0026] A plurality of self-checking points are set in the spiral channel, and the timbre of the audio in the spiral channel is recognized and marked according to the segments through the plurality of self-checking points, and the types include corresponding types and non-corresponding types;

[0027] The segment timbre marking type and the number of marked types of the audio in the spiral channel are obtained to obtain the self-check index. The calculation formula of the self-check index is:

[0028] Among them, FX is the self-check index, Z d is the total number of self-checking points, L d L is the number of audio in the spiral channel that corresponds to the timbre of the clip. f is the number of non-corresponding types of audio in the spiral channel according to the timbre of the clip, α and β are constants greater than zero, and ∈ is the non-corresponding type weight coefficient;

[0029] The self-test index that does not meet the preset threshold is fed back to the filter point as the self-test result, and the timbre of the audio segment corresponding to the self-test result is confirmed based on the filter point. If the type after the timbre confirmation is the corresponding type, there is no need to divide the audio. If the type after the timbre confirmation is the non-corresponding type, it is necessary to obtain the filter point corresponding to the non-corresponding type as the target filter point, and transfer the non-corresponding type audio segment division in the spiral channel corresponding to the filter point to the spiral channel corresponding to the target filter point.

[0030] In a preferred embodiment, the step of extracting multiple target audios based on a neural network algorithm to obtain cough audios of the user end and counting the number of cough audios includes:

[0031] Collecting personal voice information and coughing sound information of the user end to train the neural network algorithm to obtain the trained neural network algorithm;

[0032] Multiple target audios are fed into the trained neural network algorithm to extract the cough audio of the user.

[0033] Count the cough audios from the user.

[0034] In a preferred embodiment, the step of transmitting the number of cough audios to the server as the user's usage data includes:

[0035] Upload the number of cough audio statistics to the server via the http protocol;

[0036] In the server, the cough audio is sorted and counted according to the collection time.

[0037] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0038] The present invention can complete the analysis and extraction of audio, can greatly improve the accuracy of audio division, and facilitate the subsequent extraction and recognition of the user's personal voice and cough sound through a neural network algorithm, thereby improving the accuracy of cough sound recognition and extraction, and better providing analysis data on the user's physical condition. The collection of patient cough information by this system reduces the difficulty of collecting cough data for patients with chronic respiratory diseases, and ensures the accuracy and timeliness of collected cough data. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0040] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

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

[0042] Example 1, please refer to Figure 1 As shown, the real-time cough sound counting method described in this embodiment includes the following steps:

[0043] S1. Determine a user terminal, collect and process the audio of the user terminal to obtain multiple target audios;

[0044] S2. Extract multiple target audios based on a neural network algorithm to obtain cough audios from the user, and count the number of cough audios;

[0045] S3, transmitting the number of cough audios to the server as user usage data;

[0046] As described in the above steps S1-S3, with the current air pollution, the number of patients with chronic respiratory diseases is gradually increasing. In order to accurately manage chronic respiratory diseases, cough has become an important biomarker for evaluating chronic respiratory diseases; the current cough assessment mainly includes visual analog score (VAS), cough symptom score, quality of life assessment, cough frequency monitoring and cough sensitivity detection, etc. These methods have so far relied heavily on manual measurement to determine the cough frequency, which makes it difficult to reduce the difficulty of collecting cough data of patients with chronic respiratory diseases, and it is impossible to ensure the accuracy and timeliness of the collected cough data; in this application, the analysis and extraction of audio can be completed, and the accuracy of the audio division can be greatly improved, which is convenient for the subsequent extraction and recognition of the user's personal voice and cough sound through the neural network algorithm, improves the accuracy of cough sound recognition and extraction, and better provides analysis data on the physical state of the user's personnel. The collection of patient cough information by this system reduces the difficulty of collecting cough data of patients with chronic respiratory diseases, and ensures the accuracy and timeliness of the collected cough data.

[0047] In one embodiment, the step S1 of determining the user terminal, collecting and processing the audio of the user terminal to obtain multiple target audios includes:

[0048] S11, determining user information, wherein the user information includes basic user information and basic user physical information, and obtaining a user terminal according to the user information registration;

[0049] S12, formulating a user-side audio processing tree, wherein the audio processing tree includes an audio input channel, a classification channel, and a filtering point between the audio input channel and the classification channel;

[0050] S13, collecting the audio of the user end and putting it into the audio processing tree in real time for processing to obtain multiple target audios;

[0051] As described in the above steps S11-S13, in the management of chronic respiratory diseases, cough has become an important biomarker for evaluating chronic respiratory diseases. In order to complete the monitoring of patients with chronic respiratory diseases, the present application manages the state of human diseases by counting coughs for a single user. This is a system work for an individual, and an audio collector needs to be configured for a single person. For management purposes, the user's basic information and basic physical information need to be determined, wherein the user's basic information includes information such as the user's name and age, and the user's basic physical information includes a variety of medical information related to the user's past physical aspects of breathing. After that, the user end is registered according to the user information, and then the audio of the user end needs to be collected for subsequent recognition of the cough. Before processing the audio, it is necessary to formulate an audio input channel, a classification channel, and a filter point between the audio input channel and the classification channel as an audio processing tree. The filter points all have the ability to capture audio, and can divide the audio to obtain the required target audio, and then the required target audio can be processed for cough recognition.

[0052] In one embodiment, the step S12 of formulating the user-side audio processing tree includes:

[0053] S121, formulating an audio input channel, setting a plurality of filter points in the audio input channel, the plurality of filter points being interconnected, and formulating different filter information corresponding to the plurality of filter points respectively;

[0054] S122, obtaining a classification channel by diverging through the audio input channel based on the filter point;

[0055] S123, setting spiral channels in the audio input channel corresponding to the number of filter points and in association with the filter points, wherein the plurality of spiral channels are intermittently connected;

[0056] S124, using the audio input channel, the classification channel, and the filter point between the audio input channel and the classification channel as an audio processing tree;

[0057] As described in the above steps S121-S124, first, a channel for audio transmission is formulated as an audio input channel. The audio input channel is a storage device that can continuously and repeatedly play audio and display audio features. Then, multiple filter points are set in the audio input channel. The multiple filter points are interconnected. Different filter information is formulated for the multiple filter points. The filter information here is different timbres. The filter points are formulated to process audio with different timbres respectively, which is used for processing and classification of the audio, so as to facilitate the subsequent recognition of the user's cough through the neural network algorithm. The audio input and output between the spiral channels corresponding to the user's subsequent filter points are Output, the positions of multiple filter points are different, and then the corresponding filter point radiates a classification channel on the input channel, which is used to store the audio classified by the filter point. Then, in order to better classify the audio, it is necessary to set multiple spiral channels in the audio input channel according to the number of filter points. The spiral channels and the filter points are connected, and the multiple spiral channels are interconnected. The connected positions are set at intervals. Then, the audio input channel, the classification channel, and the filter points between the audio input channel and the classification channel are used as an audio processing tree. The audio processing tree is formulated and can be used for subsequent user-end audio filtering and classification processing.

[0058] In one embodiment, the step S122 of setting spiral channels in accordance with the number of filter points in the audio input channel and in association with the filter points, wherein the plurality of spiral channels are intermittently connected, comprises:

[0059] S1221, setting a corresponding number of spiral channels in the audio input channel corresponding to the number of filter points;

[0060] S1222, establishing a one-to-one correspondence between the filtering points and the spiral channels, and connecting the spiral channels and the filtering points that have a one-to-one correspondence;

[0061] S1223, selecting a plurality of interaction points of the same number in the plurality of spiral channels respectively, and establishing an interconnection relationship between the plurality of spiral channels based on the interaction points;

[0062] As described in the above steps S1221-S1223, the audio input channel is a storage device that can continuously and repeatedly play audio and display audio features. The filtering point here is an audio filtering device such as a filter. The audio filtered out by the filtering point is then stored in the corresponding classification channel. The filtering range is divided in the audio input channel. Here, the storage represented by the audio input channel is divided into a range. The range is a spiral channel, which can increase the interactivity of the range after each division. Then, a one-to-one correspondence between the filtering point and the spiral channel is established, and the spiral channels and filtering points that have a one-to-one correspondence are connected. A plurality of interaction points of the same number are selected from the plurality of spiral channels, and an intercommunication relationship between the plurality of spiral channels is established based on the interaction points. According to the information interaction between the subsequent filtering points, the output and input of audio between the spiral channels through the interaction points can be realized, which has a good audio filtering and classification effect, can greatly improve the filtering and selection of user-side audio, and obtain multiple target audios efficiently and accurately.

[0063] In one embodiment, the step S13 of collecting the audio of the user terminal and putting it into the audio processing tree in real time for processing to obtain multiple target audios includes:

[0064] S131, collecting audio from the user end in real time and inputting it into the audio input channel, performing timbre recognition on the audio from the user end through the filter point to obtain multiple audio timbres, assigning the multiple audio timbres to the filter points one by one, and separating the audio of the audio timbres corresponding to the filter points in the spiral channel;

[0065] S132, continuously collecting the audio of the user terminal in real time and separating the audio of the user terminal according to the audio timbre;

[0066] S133, the spiral channel performs self-inspection to obtain a self-inspection result, feeds back the self-inspection result to the filtering point for confirmation, re-divides the audio according to the confirmed self-inspection result, and inputs the re-divide audio into the classification channel through the spiral channel for storage to obtain multiple target audios;

[0067] As described in the above steps S131-S133, the audio of the user end is collected in real time and input into the audio input channel. The audio of the user end here is the ambient audio collected by the audio collector of the user end and the personal audio of the user end. The audio here is complex. The audio of the user end is collected in real time and input into the audio input channel in real time. The audio in the audio input channel is identified and the audio timbre is obtained to obtain multiple audio timbres. The multiple audio timbres are respectively assigned to the filter points one by one. The audio corresponding to a single audio timbre is managed by the filter point. The filter point identifies and divides the audio of the user end, and then the audio of the audio timbre corresponding to the filter point is Separated in the spiral channel, the audio of the filter point and the corresponding audio timbre is allocated in the spiral channel of the corresponding filter point. This operation is to continuously collect the audio of the user end in real time, and continuously process the audio of the user end in real time. After the audio of the user end is separated and is in the spiral channel, the spiral channel is used for self-inspection to obtain the self-inspection result, and the self-inspection result is fed back to the filter point for confirmation. The audio is re-divided according to the self-inspection result after confirmation, and the re-divided audio is input into the classification channel through the spiral channel to store multiple target audios. The classification channel here is used to store multiple target audios after classification, which can ensure the accuracy of audio division.

[0068] In one embodiment, the spiral channel performs self-test to obtain a self-test result, feeds the self-test result back to the filter point for confirmation, and performs step S133 of re-dividing the audio according to the confirmed self-test result, including:

[0069] S1331, setting a plurality of self-checking points in the spiral channel, and using the plurality of self-checking points to respectively identify the timbre of the audio in the spiral channel according to the segments and mark the types, where the types include corresponding types and non-corresponding types;

[0070] S1332, obtaining the segment timbre marking type and the number of marked types of the audio in the spiral channel to obtain a self-check index, and the calculation formula of the self-check index is:

[0071] Among them, FX is the self-check index, Z d is the total number of self-checking points, L d L is the number of audio in the spiral channel that corresponds to the timbre of the clip. f is the number of non-corresponding types of audio in the spiral channel according to the timbre of the segment, α and β are constants greater than zero, ∈ is the non-corresponding type weight coefficient, it should be noted that L f The larger the value, the d The smaller the value, that is, the larger the FX value, the greater the inaccuracy of marking the timbre fit;

[0072] S1333, the self-test index that does not meet the preset threshold is fed back to the filter point as the self-test result, and the timbre of the audio segment corresponding to the self-test result is confirmed based on the filter point. If the type after the timbre confirmation is the corresponding type, there is no need to divide the audio. If the type after the timbre confirmation is the non-corresponding type, it is necessary to obtain the filter point corresponding to the non-corresponding type as the target filter point, and transfer the non-corresponding type audio segment in the spiral channel corresponding to the filter point to the spiral channel corresponding to the target filter point;

[0073] As described in the above steps S1331-S1333, the audio will be continuously played and the audio features will be displayed in the spiral channel. A plurality of self-check points are set in the spiral channel. The self-check point is a timbre analyzer. Setting a plurality of self-check points can perform a more accurate timbre analysis on the sound, avoiding inaccuracy in the sound division caused by sound analysis errors. Then, the audio in the spiral channel is identified and marked according to the segment through the plurality of self-check points. The type includes a corresponding type and a non-corresponding type. The corresponding type here is the timbre assigned by the filter point corresponding to the spiral channel, and vice versa. The filter point is used for the division of the sound coming in from the user end and the timbre confirmation of the segment corresponding to the timbre of the self-check index that does not meet the preset threshold. The timbre inside the spiral channel is self-checked through the self-check point. The audio in the spiral channel will be analyzed by the self-check point for timbre and marked according to the analysis result. Then, the self-check index is calculated, and the self-check index that does not meet the preset threshold is fed back to the filter point as the self-check result. The timbre of the audio segment corresponding to the self-check result is confirmed based on the filter point. If If the type after the timbre confirmation is a corresponding type, there is no need to divide the audio. If the type after the timbre confirmation is a non-corresponding type, it is necessary to obtain the filter point corresponding to the non-corresponding type as the target filter point, and transfer the segment division of the non-corresponding type audio in the spiral channel corresponding to the filter point to the spiral channel corresponding to the target filter point. Since the filter points are interconnected, when there is a non-corresponding type of audio in the current filter point, the filter point corresponding to the non-corresponding type of audio can be obtained through the interconnected relationship, and then the segment division of the non-corresponding type of audio in the spiral channel corresponding to the filter point is transferred to the spiral channel corresponding to the target filter point, which can complete the analysis and extraction of the audio, and can greatly improve the accuracy of the audio division, which is convenient for the subsequent extraction and recognition of the user's personal voice and cough sound through the neural network algorithm, improves the accuracy of cough sound recognition and extraction, and better provides analysis data on the physical state of the user. The collection of patient cough information by this system reduces the difficulty of collecting cough data of patients with chronic respiratory diseases, and ensures the accuracy and timeliness of collected cough data.

[0074] In one embodiment, the step S2 of extracting the cough audios of the user terminal by respectively extracting the plurality of target audios based on the neural network algorithm and counting the number of the cough audios comprises:

[0075] S21, collecting personal voice information of the user and coughing sound information to train the neural network algorithm to obtain a trained neural network algorithm;

[0076] S22, inputting the multiple target audios into the trained neural network algorithm to extract the cough audio of the user end;

[0077] S23, counting the number of coughing audios from the user end;

[0078] As described in the above steps S21-S23, if the user's personal cough in the target audio is to be extracted by the neural network algorithm, the neural network algorithm needs to be trained here, and the user's personal voice information, including the timbre of the voice, is used to train the neural network algorithm, with the purpose of identifying the user's personal voice in multiple target audios, and then identifying and extracting the cough audio on the basis of identifying the user's personal voice. The sound features of the cough audio are additionally trained for the neural network algorithm, so that the trained neural network algorithm can not only identify the user's personal voice but also extract the cough audio on this basis. The cough audio here is a divided cough audio, and there are as many coughs as there are cough audios. The training of the neural network algorithm here is performed using a single cough, and then the cough audio of the user can be counted, which is convenient for subsequent quantity storage and understanding of the cough status of the user.

[0079] In one embodiment, the step S3 of transmitting the number of cough audios to the server as the user's usage data includes:

[0080] S31, uploading the number of cough audio statistics to the server via the http protocol;

[0081] S32. Sort and count the number of cough audios in the server according to the collection time;

[0082] As described in the above steps S31 and S32, after the front-end system recognizes and extracts the cough audio of the user, in order to manage the user data, the number of cough audio statistics needs to be uploaded to the server through the http protocol. In order to manage the cough cycle and time of the user, it is necessary to sort the statistics according to the time of collecting the cough audio and store them in the server, so that the cough situation of the user can be understood more clearly and better assist the subsequent physical treatment of the user.

[0083] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A real-time cough sound counting method, characterized in that: The following steps are involved: Determine a user terminal, collect and process the audio of the user terminal to obtain multiple target audios; Based on the neural network algorithm, multiple target audios are extracted to obtain the cough audios of the user end, and the number of cough audios is counted; The number of cough audios is transmitted to the server as the user's usage data.

2. A real-time cough sound counting method according to claim 1, characterized in that: The step of determining the user terminal, collecting and processing the audio of the user terminal to obtain multiple target audios includes: Determine user information, wherein the user information includes basic user information and basic user physical information, and obtain a user terminal according to the user information registration; Formulate a user-side audio processing tree, wherein the audio processing tree includes an audio input channel, a classification channel, and a filtering point between the audio input channel and the classification channel; The audio from the user end is collected and put into the audio processing tree in real time for processing to obtain multiple target audios.

3. A real-time cough sound counting method according to claim 2, characterized in that: The step of formulating the user-side audio processing tree includes: Formulate an audio input channel, set multiple filter points in the audio input channel, the multiple filter points are interconnected, and different filter information is formulated corresponding to the multiple filter points; Based on the filtering point, the classification channel is obtained by diverging through the audio input channel; In the audio input channel, a spiral channel is set corresponding to the number of filter points and associated with the filter points, and the plurality of spiral channels are intercommunicated with each other at intervals; The audio input channel, the classification channel, and the filtering points between the audio input channel and the classification channel are used as an audio processing tree.

4. A real-time cough sound counting method according to claim 3, characterized in that: The step of setting spiral channels in the audio input channel in accordance with the number of filter points and in association with the filter points, and intermittently communicating with each other between the plurality of spiral channels, comprises: Set a corresponding number of spiral channels in the audio input channel corresponding to the number of filter points; Establish a one-to-one correspondence between the filtering points and the spiral channels, and connect the spiral channels and the filtering points that have a one-to-one correspondence; A plurality of interaction points of the same number are respectively selected in the plurality of spiral channels, and an interconnection relationship between the plurality of spiral channels is established based on the interaction points.

5. A real-time cough sound counting method according to claim 4, characterized in that: The step of collecting the audio of the user end and putting it into the audio processing tree in real time for processing to obtain multiple target audios includes: The audio of the user end is collected in real time and input into the audio input channel, the timbre of the audio of the user end is recognized through the filter point to obtain multiple audio timbres, the multiple audio timbres are respectively assigned to the filter points, and the audio of the audio timbre corresponding to the filter point is separated in the spiral channel; Continuously collect the user's audio in real time and separate the user's audio according to the audio timbre; The spiral channel performs self-inspection to obtain a self-inspection result, and the self-inspection result is fed back to the filter point for confirmation. The audio is re-divided according to the confirmed self-inspection result, and the re-divided audio is input into the classification channel through the spiral channel to store multiple target audios.

6. A real-time cough sound counting method according to claim 1, characterized in that: The spiral channel performs self-test to obtain a self-test result, feeds the self-test result back to the filtering point for confirmation, and performs the step of re-dividing the audio according to the confirmed self-test result, including: A plurality of self-checking points are set in the spiral channel, and the timbre of the audio in the spiral channel is recognized and marked according to the segments through the plurality of self-checking points, and the types include corresponding types and non-corresponding types; The segment timbre marking type and the number of marked types of the audio in the spiral channel are obtained to obtain the self-check index. The calculation formula of the self-check index is: Among them, FX is the self-check index, Z d is the total number of self-checking points, L d L is the number of audio in the spiral channel that corresponds to the timbre of the clip. f is the number of non-corresponding types of audio in the spiral channel according to the timbre of the clip, α and β are constants greater than zero, and ∈ is the non-corresponding type weight coefficient; The self-test index that does not meet the preset threshold is fed back to the filter point as the self-test result, and the timbre of the audio segment corresponding to the self-test result is confirmed based on the filter point. If the type after the timbre confirmation is the corresponding type, there is no need to divide the audio. If the type after the timbre confirmation is the non-corresponding type, it is necessary to obtain the filter point corresponding to the non-corresponding type as the target filter point, and transfer the non-corresponding type audio segment division in the spiral channel corresponding to the filter point to the spiral channel corresponding to the target filter point.

7. A real-time cough sound counting method according to claim 5, characterized in that: The step of extracting multiple target audios based on the neural network algorithm to obtain cough audios from the user end and counting the number of cough audios includes: Collecting personal voice information of the user and coughing sound information to train the neural network algorithm to obtain the trained neural network algorithm; Multiple target audios are fed into the trained neural network algorithm to extract the cough audio of the user. Count the cough audios from the user.

8. A real-time cough sound counting method according to claim 1, characterized in that: The step of transmitting the number of cough audios to the server as the user's usage data includes: Upload the number of cough audio statistics to the server via the http protocol; In the server, the cough audio is sorted and counted according to the collection time.

Images