Personalized voice rehabilitation training method and system based on deep learning

Abstract

The invention discloses a personalized voice rehabilitation training method and system based on deep learning, belongs to the technical field of audio data recognition, and can automatically analyze voice characteristics of a patient, customize a training scheme and improve a rehabilitation effect. Comprising the following steps: acquiring a training sample set; training the deep learning model according to the training sample set to obtain a voice rehabilitation model; acquiring first audio data and inquiry information of a first user; based on the first audio data and the inquiry information, generating a first rehabilitation training scheme through a voice rehabilitation model; acquiring second audio data of the first user; the second audio data is audio data after the first user performs rehabilitation training according to the first rehabilitation training scheme and reaches the first preset training time; and based on the second audio data, adjusting the first rehabilitation training scheme through the voice rehabilitation model to generate a second rehabilitation training scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of audio data recognition, and in particular to a personalized voice rehabilitation training method and system based on deep learning. Background Art

[0002] Sound is the carrier of human language communication. The vocal system produces sound driven by the vibration of the vocal cords, and it is transmitted through the channel composed of the throat and mouth. Sound carries different information and is an indispensable means of communication between people. The sound produced by the vocal system can be described by a waveform signal, which is called voice. When the vocal organs are in a normal state, the vibration of the vocal cords has an obvious periodic pattern, and the transmission channel composed of the throat and mouth also changes regularly, so the voice produced is also very regular. Patients with damaged vocal cords need voice rehabilitation to help them restore or improve their voice function.

[0003] Existing voice rehabilitation training programs usually adopt standardized methods. Although they can meet the needs of some patients to a certain extent, these programs often fail to fully consider the unique voice characteristics and individualized rehabilitation needs of each patient. For example, different patients have significant differences in the structure, function and pathological state of the vocal cords. A unified training program may lead to poor individual rehabilitation effects, thereby prolonging the rehabilitation period.

[0004] In addition, current voice rehabilitation analysis methods mainly rely on manual evaluation and guidance from professional therapists. This human-dependent evaluation method faces the problems of subjectivity and lack of consistency. Different therapists may come to different conclusions when evaluating the same patient due to differences in personal experience and professional level, resulting in a lack of objectivity and consistency in the training program.

[0005] The disclosure of the above background technology content is only used to assist in understanding the concept and technical solution of the present invention. It does not necessarily belong to the prior art of this patent application. In the absence of clear evidence that the above content has been disclosed on the filing date of this patent application, the above background technology should not be used to evaluate the novelty and creativity of the present application. Summary of the invention

[0006] The present application provides a personalized voice rehabilitation training method and system based on deep learning, which can automatically analyze the patient's voice characteristics, tailor training plans, and improve rehabilitation effects.

[0007] To achieve the above objectives, the present application discloses the following technical solutions:

[0008] In a first aspect, the present application provides a personalized voice rehabilitation training method based on deep learning, comprising the following steps:

[0009] Get a training sample set;

[0010] The deep learning model is trained according to the training sample set to obtain a voice rehabilitation model;

[0011] Acquiring first audio data and medical consultation information of a first user;

[0012] Based on the first audio data and the consultation information, a first rehabilitation training program is generated through a voice rehabilitation model;

[0013] Acquire second audio data of the first user; the second audio data is audio data of the first user after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time;

[0014] Based on the second audio data, the first rehabilitation training program is adjusted through the voice rehabilitation model to generate a second rehabilitation training program.

[0015] In the embodiments of the present application, by training the deep learning model with a training sample set, a voice rehabilitation model capable of identifying and analyzing the unique voice characteristics of the patient and generating a personalized rehabilitation training program that is highly matched with the patient's voice characteristics and health condition can be obtained, which effectively solves the problem of lack of personalization of the rehabilitation training program provided by the prior art;

[0016] In addition, the generated rehabilitation training program is based on the objective analysis results of the model rather than manual evaluation, which can eliminate the subjectivity in manual evaluation and make the rehabilitation training program more objective and consistent. In addition, the rehabilitation program can be dynamically adjusted by obtaining the patient's audio data during the training process. In this way, on the one hand, it can ensure that the rehabilitation training program is personalized both in the initial stage and during the training process; on the other hand, the rehabilitation training program can adapt to the patient's recovery in real time during the rehabilitation process, ensuring that the user's training plan can always adapt to their current voice condition, avoiding the weakening or stagnation of the training effect due to the long-term use of a fixed program, thereby ensuring the scientificity and effectiveness of the rehabilitation process.

[0017] In some possible implementations of the first aspect, the training sample set includes historical audio data of multiple users, corresponding consultation information, and related rehabilitation result data; the consultation information is recorded by a professional physician, and the consultation information includes the user's medical history, symptom description, diagnosis result, and recommended treatment plan; the rehabilitation result data includes follow-up audio data of the user after completing the recommended treatment plan and rehabilitation result data corresponding to the audio data;

[0018] The steps of training the deep learning model according to the training sample set to obtain the voice rehabilitation model include:

[0019] Preprocess each user's historical audio data;

[0020] Label the audio features with the corresponding medical consultation information and rehabilitation result data to form one-to-one corresponding training samples;

[0021] Multiple training samples are input into the deep learning model for training to obtain a voice rehabilitation model. In this way, the sample data used for training is more comprehensive, which can improve the generalization ability of the voice rehabilitation model, so that the voice rehabilitation model can better adapt to various situations when facing different patients, reduce the risk of overfitting specific samples, and enable the voice rehabilitation model to more accurately provide rehabilitation training programs that meet the specific needs of patients and improve the treatment effect. In addition, comprehensive sample data can also enhance the robustness of the voice rehabilitation model, so that it can learn to deal with complex situations during the training process, so that the model can show stronger stability in practical applications.

[0022] In addition, by labeling the audio features with the corresponding medical consultation information and rehabilitation results data to form one-to-one corresponding training samples, the deep learning model can more effectively learn and extract the complex relationship between audio features and voice rehabilitation effects, which is conducive to improving the accuracy and efficiency of model training.

[0023] In some possible implementations of the first aspect, the steps of generating the first rehabilitation training program and / or the second rehabilitation training program further include:

[0024] Get the target training days of the first user;

[0025] Based on the first user's medical history, symptom characteristics, diagnosis results and medical guidance standards, the voice rehabilitation model is used to determine the reasonable range of training days;

[0026] Compare and analyze the target training days with the reasonable training days range determined by the voice rehabilitation model;

[0027] When the target training days are within the reasonable training days range, a rehabilitation training plan of corresponding duration is generated based on the target training days;

[0028] When the target training days are less than the minimum value of the reasonable training days range, a rehabilitation training plan is generated based on the minimum value of the reasonable training days range, and a prompt message is output to the user to point out the unreasonableness of the target training days.

[0029] In this way, the voice rehabilitation model can be used to analyze the patient's health status and medical history, determine a reasonable range of training days, and help patients set a scientific and achievable rehabilitation goal, avoiding the risk of rehabilitation plan failure or overtraining due to blind pursuit of goals.

[0030] In addition, patients can be allowed to set their own training goals and adjust them according to the recommendations of the voice rehabilitation model, which enhances the patient's sense of participation and initiative in the rehabilitation process. When patients see that their goals are consistent with the model's recommendations, they will have more trust in the rehabilitation program and will be more actively involved in the training, which not only improves the effectiveness of the treatment, but also enhances the patient's satisfaction and confidence.

[0031] In addition, by comparing the target training days set by the patient with the reasonable range recommended by the model, on the one hand, a rehabilitation plan that is more in line with the patient's actual situation can be generated, ensuring that the patient can maintain an appropriate training intensity during the rehabilitation process, which helps to improve the overall rehabilitation effect and enable the patient to achieve the best rehabilitation state in a shorter period of time, reducing the risks that may arise during the rehabilitation process, especially the risk of voice damage caused by overtraining or improper training; on the other hand, it can ensure that the second rehabilitation training plan will not exceed the target training days initially set, ensuring the continuity and consistency of the rehabilitation process, so that patients can make steady progress within a scientific training framework.

[0032] In some possible implementations of the first aspect, the step of adjusting the first rehabilitation training program based on the second audio data and using the voice rehabilitation model to generate the second rehabilitation training program includes:

[0033] Converting the first audio data and the second audio data into a feature vector that can be used for model analysis by using a feature extraction algorithm;

[0034] Comparing the feature vector of the second audio data with the feature vector of the first audio data, and calculating a difference index between the two;

[0035] According to the difference indicators, the rehabilitation progress is obtained;

[0036] The training parameters of the first rehabilitation training program are adjusted according to the rehabilitation progress to obtain the second rehabilitation training program. In this way, the timeliness and accuracy of the rehabilitation effect can be improved.

[0037] In some possible implementations of the first aspect, the step of generating a first rehabilitation training program based on the first audio data and the consultation information and using a voice rehabilitation model includes:

[0038] Preprocessing the first audio data to extract audio features including volume, pitch, formant, duration, and speech speed;

[0039] Perform semantic analysis on the symptom description, medical history information, diagnosis results, etc. in the medical consultation information to extract the corresponding feature labels;

[0040] Associating the audio features with the feature labels of the medical consultation information and inputting them into the voice rehabilitation model as input data;

[0041] The voice rehabilitation model analyzes the user's voice status based on the input audio features and the feature labels of the medical consultation information to obtain the analysis results;

[0042] According to the analysis results, a first rehabilitation training program is generated, so that the accuracy and pertinence of program generation can be improved, making individualized rehabilitation training more effective.

[0043] In some possible implementations of the first aspect, the personalized voice rehabilitation training method based on deep learning further includes the following steps:

[0044] Calculating the similarity index between the first rehabilitation training program of the first user and the first rehabilitation training programs of other users through a feature similarity analysis algorithm;

[0045] Based on the similarity index, users whose similarity with the first user is higher than the preset threshold are screened out to form a set of similar users. In this way, on the one hand, the reference basis of personalized training programs can be enhanced, and the successful programs of others can be learned to improve the rehabilitation effect; on the other hand, it is convenient to select rehabilitation partners to ensure that the rehabilitation partners have a high degree of relevance in voice characteristics, rehabilitation needs and training programs, promote more effective communication between partners, and enhance mutual support; establish a positive competitive atmosphere to further motivate users. In a partnership with common goals and similar rehabilitation paths, users are often motivated by the progress of their partners, prompting themselves to train harder. This benign competitive relationship can increase the user's training enthusiasm and thus improve the overall rehabilitation effect.

[0046] In addition, it can also provide users with a platform for mutual supervision. During the rehabilitation process, users can observe and evaluate each other's progress, give timely feedback and suggestions, and thus form a positive supervision mechanism. This supervision not only enhances the user's sense of responsibility during the training process, but also reduces slackness and interruptions in training, making the rehabilitation process more continuous and stable.

[0047] In addition, through this partnership, users can share successful experiences and strategies for coping with challenges. This experience sharing is not only an encouragement to individuals, but also a positive promotion for the entire rehabilitation process. When users see successful cases of similar situations, they tend to increase their confidence and be more determined to implement rehabilitation training programs.

[0048] In some possible implementations of the first aspect, the personalized voice rehabilitation training method based on deep learning further includes the following steps:

[0049] Generate a corresponding recommended rehabilitation training plan according to the rehabilitation training plan of one of the users in the similar user set;

[0050] The recommended rehabilitation training program is used as input, combined with the second audio data, and the first rehabilitation training program is adjusted through the voice rehabilitation model to obtain the second rehabilitation training program. In this way, it can ensure that the training progress of the current user is synchronized with the training progress of his rehabilitation partner. This can help maintain and enhance the partnership; on the other hand, it can improve the effectiveness of rehabilitation, allowing patients to psychologically feel the common progress with their rehabilitation partners on the road to rehabilitation, thereby enhancing the enthusiasm and persistence of treatment.

[0051] In a second aspect, the embodiment of the present application provides a personalized voice rehabilitation training system based on deep learning, including:

[0052] A first acquisition module is used to acquire a training sample set;

[0053] The first building module is used to train the deep learning model according to the training sample set to obtain a voice rehabilitation model;

[0054] A second acquisition module is used to acquire the first audio data and medical consultation information of the first user;

[0055] A first generating module, configured to generate a first rehabilitation training program based on the first audio data and the consultation information and through a voice rehabilitation model;

[0056] The third acquisition module is used to acquire second audio data of the first user; the second audio data is audio data after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time;

[0057] The second generating module is used to adjust the first rehabilitation training program based on the second audio data and through the voice rehabilitation model to generate a second rehabilitation training program.

[0058] In some possible implementations of the second aspect, the training sample set includes historical audio data of multiple users, corresponding consultation information, and related rehabilitation result data; the consultation information is recorded by a professional physician, and the consultation information includes the user's medical history, symptom description, diagnosis result, and recommended treatment plan; the rehabilitation result data includes follow-up audio data of the user after completing the recommended treatment plan and rehabilitation result data corresponding to the audio data;

[0059] The first building block is specifically used for:

[0060] Preprocess each user's historical audio data;

[0061] Label the preprocessed audio features and the corresponding medical consultation information and rehabilitation result data to form one-to-one corresponding training samples;

[0062] Multiple training samples are input into the deep learning model for training to obtain a voice rehabilitation model.

[0063] In some possible implementations of the second aspect, the first generating module and / or the second generating module is further configured to:

[0064] Get the target training days of the first user;

[0065] Based on the first user's medical history, symptom characteristics, diagnosis results and medical guidance standards, the voice rehabilitation model is used to determine the reasonable range of training days;

[0066] Compare and analyze the target training days with the reasonable training days range determined by the voice rehabilitation model;

[0067] When the target training days are within the reasonable training days range, a rehabilitation training plan of corresponding duration is generated based on the target training days;

[0068] When the target training days are less than the minimum value of the reasonable training days range, a rehabilitation training plan is generated based on the minimum value of the reasonable training days range, and a prompt message is output to the user to point out the unreasonableness of the target training days.

[0069] In some possible implementations of the second aspect, the second generating module is further specifically used for:

[0070] Converting the first audio data and the second audio data into a feature vector that can be used for model analysis by using a feature extraction algorithm;

[0071] Comparing the feature vector of the second audio data with the feature vector of the first audio data, and calculating a difference index between the two;

[0072] According to the difference indicators, the rehabilitation progress is obtained;

[0073] The training parameters of the first rehabilitation training program are adjusted according to the rehabilitation progress to obtain a second rehabilitation training program.

[0074] In some possible implementations of the second aspect, the first generating module is further specifically used for:

[0075] Preprocessing the first audio data to extract audio features including volume, pitch, formant, duration, and speech speed;

[0076] Perform semantic analysis on the symptom description, medical history information, and diagnosis results in the medical consultation information to extract the corresponding feature labels;

[0077] Associating the audio features with the feature labels of the medical consultation information and inputting them into the voice rehabilitation model as input data;

[0078] The voice rehabilitation model analyzes the user's voice status based on the input audio features and the feature labels of the medical consultation information to obtain the analysis results;

[0079] According to the analysis results, a first rehabilitation training program is generated.

[0080] In some possible implementations of the second aspect, the personalized voice rehabilitation training system based on deep learning further includes a first screening module, which is used to calculate a similarity index between the first rehabilitation training program of the first user and the first rehabilitation training programs of other users through a feature similarity analysis algorithm;

[0081] Based on the similarity index, users whose similarity with the first user is higher than a preset threshold are screened out to form a similar user set.

[0082] In some possible implementations of the second aspect, the second generating module is further specifically used for:

[0083] Generate a corresponding recommended rehabilitation training plan according to the rehabilitation training plan of one of the users in the similar user set;

[0084] The recommended rehabilitation training program is used as input, combined with the second audio data, and the first rehabilitation training program is adjusted through the voice rehabilitation model to obtain the second rehabilitation training program.

[0085] In a third aspect, an embodiment of the present application provides a terminal device, including one or more processors; a storage device, on which one or more programs are stored;

[0086] When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any technical solution of the first aspect.

[0087] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, a method such as any one of the technical solutions in the first aspect is implemented.

[0088] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method of any one of the technical solutions of the first aspect.

[0089] Among them, the technical effects brought about by any design method in the second to fifth aspects can refer to the technical effects brought about by different design methods in the first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0091] Figure 1 A flowchart of a personalized voice rehabilitation training method based on deep learning provided in some embodiments of the present application;

[0092] Figure 2 A schematic diagram of the structure of a personalized voice rehabilitation training system based on deep learning provided in some embodiments of the present application;

[0093] Figure 3 A schematic diagram of the structure of a terminal device suitable for implementing some embodiments of the present application. DETAILED DESCRIPTION

[0094] Specific embodiments of the present invention will now be mentioned in detail. Although the present invention is described in conjunction with these specific embodiments, it should be appreciated that it is not intended to limit the present invention to these specific embodiments. On the contrary, these embodiments are intended to cover substitutions, changes or equivalent embodiments that may be included in the spirit and scope of the invention defined by the claims. In the following description, a large number of specific details are set forth in order to provide a comprehensive understanding of the present invention. The present invention may be implemented without some or all of these specific details.

[0095] When used in conjunction with "including," "methods comprising," or similar language in this specification and the appended claims, the singular forms "a," "an," and "the" include plural references unless the context clearly dictates otherwise. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0096] Sound is the carrier of human language communication. The vocal system produces sound driven by the vibration of the vocal cords, and it is transmitted through the channel composed of the throat and mouth. Sound carries different information and is an indispensable means of communication between people. The sound produced by the vocal system can be described by a waveform signal, which is called voice. When the vocal organs are in a normal state, the vibration of the vocal cords has an obvious periodic pattern, and the transmission channel composed of the throat and mouth also changes regularly, so the voice produced is also very regular. Patients with damaged vocal cords need voice rehabilitation to help them restore or improve their voice function.

[0097] Existing voice rehabilitation training programs usually adopt standardized methods. Although they can meet the needs of some patients to a certain extent, these programs often fail to fully consider the unique voice characteristics and individualized rehabilitation needs of each patient. For example, different patients have significant differences in the structure, function and pathological state of the vocal cords. A unified training program may lead to poor individual rehabilitation effects, thereby prolonging the rehabilitation period.

[0098] In addition, current voice rehabilitation analysis methods mainly rely on manual evaluation and guidance from professional therapists. This human-dependent evaluation method faces the problems of subjectivity and lack of consistency. Different therapists may come to different conclusions when evaluating the same patient due to differences in personal experience and professional level, resulting in a lack of objectivity and consistency in the training program.

[0099] In response to the above technical problems, the overall idea of ​​the technical solution provided by this application is as follows: obtain a training sample set; train a deep learning model according to the training sample set to obtain a voice rehabilitation model; obtain first audio data and consultation information of a first user; based on the first audio data and consultation information, generate a first rehabilitation training plan through the voice rehabilitation model; obtain second audio data of the first user; the second audio data is the audio data of the first user after the first rehabilitation training plan is completed and the first preset training time is reached; based on the second audio data, the first rehabilitation training plan is adjusted through the voice rehabilitation model to generate a second rehabilitation training plan.

[0100] The above technical solution, by using a training sample set to train a deep learning model, can obtain a voice rehabilitation model that has the ability to identify and analyze the patient's unique voice characteristics and generate a personalized rehabilitation training program that is highly matched with the patient's voice characteristics and health conditions, effectively solving the problem of lack of personalization in the rehabilitation training programs provided by the prior art;

[0101] In addition, the generated rehabilitation training program is based on the objective analysis results of the model rather than manual evaluation, which can eliminate the subjectivity in manual evaluation and make the rehabilitation training program more objective and consistent. In addition, the rehabilitation program can be dynamically adjusted by obtaining the audio data of the patient during the training process. In this way, on the one hand, it can ensure that the rehabilitation training program is personalized both in the initial stage and during the training process; on the other hand, the rehabilitation training program can adapt to the patient's recovery in real time during the rehabilitation process, ensuring that the user's training plan can always adapt to his current voice condition, avoiding the weakening or stagnation of the training effect caused by the long-term use of a fixed program, thereby ensuring the scientificity and effect of the rehabilitation process. In addition, by labeling the audio features with the corresponding consultation information and rehabilitation result data to form a one-to-one corresponding training sample, the deep learning model can more effectively learn and extract the complex relationship between audio features and voice rehabilitation effects, which is conducive to improving the accuracy and efficiency of model training.

[0102] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0103] See also Figure 1 The present application embodiment provides a personalized voice rehabilitation training method based on deep learning, comprising the following steps:

[0104] S101: Obtain a training sample set.

[0105] In some embodiments, the executor (e.g., a computing device) of the personalized voice rehabilitation training method based on deep learning can obtain a set of training samples from the electronic medical record system (EMR) of a hospital or rehabilitation center, the data management system of a speech rehabilitation specialist, or the patient's follow-up record library or self-recording through a wired or wireless connection. The training sample set specifically includes historical audio data of multiple users, corresponding medical consultation information, and related rehabilitation result data; each audio data in the historical audio data represents the voice condition of its corresponding patient at different time points; the audio data should be recorded under standardized conditions, specifically including but not limited to vocal samples recorded using a calibrated microphone in a quiet environment, and the samples should cover different vocalization situations such as normal vocalization, voice fatigue, and pitch testing. The audio data of each user should cover different time points in multiple follow-ups to ensure that the recovery process of the vocalization state is covered.

[0106] The consultation information is recorded by professional physicians and includes the user's medical history, symptom description, diagnosis results, and recommended treatment plans. The consultation information is digitized and natural language processing (NLP) technology is used to extract key features, such as symptom keywords, diagnosis categories, and types of treatment plans, to form structured data for subsequent model training.

[0107] The rehabilitation result data includes the follow-up audio data of the user after completing the recommended treatment plan and the rehabilitation result data corresponding to the audio data. The rehabilitation result data includes the sound quality recorded at each follow-up, the patient's subjective feelings (recorded through questionnaires or rating scales), and the doctor's objective evaluation results (such as the degree of improvement in the voice assessed through GRBAS scores). These data should be structured to form feature labels that can be used for model learning. The evaluation of rehabilitation effects should include short-term effects (at the end of treatment) and long-term effects (the specific time interval depends on actual needs, for example, follow-up data 3 months and 6 months after the end of treatment), and should be quantified using unified standards (such as pitch stability, speech clarity, pronunciation persistence, etc.).

[0108] S102: training the deep learning model according to the training sample set to obtain a voice rehabilitation model;

[0109] Specifically, in some embodiments, the execution subject may train the deep learning model according to the training sample set through the following steps to obtain the voice rehabilitation model:

[0110] The first step is to preprocess each user’s historical audio data;

[0111] Specifically, it includes removing background noise (using noise suppression algorithms, such as spectral subtraction), signal enhancement (such as normalization), and audio slicing (splitting long-duration audio into segments of fixed length). In terms of feature extraction, it is necessary to extract multiple features of the audio signal, which can be frequency domain features: including Mel-frequency cepstral coefficients (MFCC), linear prediction coefficients (LPC), cepstral features, etc. These features help capture the spectral information of the sound. Time domain features: such as short-time energy and zero-crossing rate, are used to capture the time-varying characteristics of speech signals. Time-frequency domain features: such as short-time Fourier transform (STFT) and continuous wavelet transform (CWT), these features are used to analyze the changes in the frequency of speech signals over time.

[0112] The second step is to label the audio features with the corresponding medical information and rehabilitation results data to form one-to-one corresponding training samples;

[0113] Specifically, NLP technology can be used to process text-based medical information, extract structured labels (such as disease classification, symptom level), and encode these labels for input into the deep learning model. Labeling of rehabilitation outcome data can include standardizing quantitative indicators (such as sound quality scores, rehabilitation cycles) to make them suitable for model input, and classifying qualitative feedback to generate appropriate classification labels.

[0114] The third step is to input multiple training samples into the deep learning model for training to obtain a voice rehabilitation model.

[0115] Specifically, the spatial features in the audio data can be extracted through a convolutional neural network (CNN) to process the spectral image of the audio signal. The time series characteristics of the audio data are processed through a recurrent neural network (RNN) or a long short-term memory network (LSTM) to help the model capture the dynamic changes of the speech signal. A hybrid architecture combining CNN and RNN (such as CRNN) simultaneously processes the spatial and temporal characteristics of the audio signal. The model training adopts a supervised learning method, using the Adam optimizer or the stochastic gradient descent (SGD) optimizer, and gradually optimizes the model parameters (such as the convolution kernel size, the number of RNN units) through multiple iterations of training and evaluation of the validation set. Preferably, data enhancement techniques can also be used to expand the training data set, including random editing of audio, speed adjustment, noise addition, etc., to improve the generalization ability of the model.

[0116] The cross-validation method is used to ensure the consistency of the model's performance on different sample sets and avoid overfitting. During the training and validation process, the learning rate, batch size, and regularization parameters are dynamically adjusted, and finally evaluated on the test set to verify the actual application effect of the model. The evaluation indicators include the model's accuracy, recall rate, F1 score, mean absolute error (MAE), etc., which are used to quantify the accuracy and stability of the model's prediction of the effect of voice rehabilitation.

[0117] In this way, the sample data used for training is more comprehensive, which can improve the generalization ability of the voice rehabilitation model, so that the voice rehabilitation model can better adapt to various situations when facing different patients, reduce the risk of overfitting specific samples, and enable the voice rehabilitation model to more accurately provide rehabilitation training programs that meet the specific needs of patients and improve the treatment effect. In addition, comprehensive sample data can also enhance the robustness of the voice rehabilitation model, so that it can learn to deal with complex situations during the training process, so that the model can show stronger stability in practical applications.

[0118] S103: Acquire first audio data and medical consultation information of the first user;

[0119] Specifically, the above-mentioned execution entity can obtain the first audio data and consultation information of the first user from the electronic medical record system of the hospital or rehabilitation center through a wired connection or a wireless connection: It can be understood that in other embodiments, the first audio data can also be recorded by the user himself.

[0120] S104: Generate a first rehabilitation training program based on the first audio data and the consultation information and through a voice rehabilitation model.

[0121] Specifically, the execution subject may generate a first rehabilitation training program based on the first audio data and the consultation information and through the voice rehabilitation model through the following steps:

[0122] The first step is to pre-process the first audio data to extract audio features including volume, pitch, formant, duration and speech speed;

[0123] The second step is to perform semantic analysis on the symptom description, medical history information, diagnosis results, etc. in the medical consultation information to extract the corresponding feature labels;

[0124] Specifically, natural language processing (NLP) technology can be used to perform semantic analysis and extract relevant feature tags.

[0125] The third step is to associate the audio features with the feature labels of the medical consultation information and input them into the voice rehabilitation model as input data;

[0126] In the fourth step, the voice rehabilitation model analyzes the user's voice status based on the input audio features and the feature labels of the medical consultation information to obtain the analysis results;

[0127] Specifically, deep learning algorithms (such as CNN, RNN, etc.) can be used to understand the user's voice status, identify the user's voice problems and their severity, and output analysis results. The output analysis results include the current status of the user's voice (such as normal, slightly damaged, severely damaged), corresponding suggestions, and possible causes of voice problems (such as improper vocalization, vocal cord fatigue, etc.).

[0128] The fifth step is to generate the first rehabilitation training program based on the analysis results.

[0129] Specifically, a personalized first rehabilitation training program is generated based on the user's specific circumstances (such as age, gender, occupation, living habits, etc.). The rehabilitation training program includes but is not limited to voice training (such as pronunciation exercises, breathing control), voice protection measures (such as paying attention to the vocal environment) and daily life suggestions (such as maintaining a good work and rest schedule). In this way, the accuracy and pertinence of program generation can be improved, making individualized rehabilitation training more effective.

[0130] S105: Acquire second audio data of the first user; the second audio data is audio data of the first user after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time;

[0131] Specifically, the execution subject may connect to the recording device via a wired connection or a wireless connection to obtain the second audio data. The second audio data may be actual audio data, or second audio data after a preset time interval.

[0132] S106: Based on the second audio data, the first rehabilitation training program is adjusted through the voice rehabilitation model to generate a second rehabilitation training program.

[0133] The execution subject may generate a second rehabilitation training program by adjusting the first rehabilitation training program based on the second audio data and the voice rehabilitation model through the following steps:

[0134] The first step is to convert the first audio data and the second audio data into a feature vector that can be used for model analysis through a feature extraction algorithm;

[0135] Specifically, the time domain signal can be converted into a frequency domain signal by performing a short-time Fourier transform (STFT) on each frame of audio to obtain a spectrogram of each frame of audio. The spectrogram is mapped to the Mel frequency scale to simulate the auditory characteristics of the human ear. The cepstral coefficients are calculated as the feature vector of the audio. The fundamental frequency of each frame of audio, that is, the pitch value, is calculated using an autocorrelation or harmonic formant detection method to represent the frequency characteristics of the sound. The pitch values ​​of multiple frames are averaged to generate a pitch feature vector for the entire audio sample. The loudness of the audio is calculated by calculating the energy of each frame of audio signal or by weighting using an A-characteristic curve. These loudness values ​​are smoothed to generate a sound intensity feature vector. The linear predictive coding (LPC) method or the formant analysis method is used to extract the formant frequencies in the audio, especially the first two formants (F1 and F2), which represent the resonance characteristics of the sound. The formant frequencies are used as components of the feature vector. The various feature vectors of each frame of audio are spliced ​​to form a complete frame-level feature vector.

[0136] The feature vectors of all frames are summarized, and the frame-level feature vectors can be reduced in dimension or summarized into feature vectors for model analysis by methods such as averaging, weighted averaging or principal component analysis (PCA).

[0137] In the second step, the feature vector of the second audio data is compared with the feature vector of the first audio data, and a difference index between the two is calculated.

[0138] Specifically, the difference index may include but is not limited to: mean square error is used to evaluate the average difference between two signals. Dynamic time warping is used to compare the time difference between two audio segments, which is applicable to changes in rhythm or speed in audio. Correlation coefficient is used to measure the linear relationship between two feature vectors.

[0139] The third step is to obtain the rehabilitation progress based on the difference indicators.

[0140] Specifically, the difference index is passed as input to the voice rehabilitation model, which should have been pre-trained to identify and evaluate the characteristics of rehabilitation progress. The evaluation results output by the voice rehabilitation model can cover the user's rehabilitation progress, including the degree of improvement in voice quality, changes in pronunciation ability, etc.

[0141] The fourth step is to adjust the training parameters of the first rehabilitation training program according to the progress of rehabilitation to obtain a second rehabilitation training program.

[0142] Specifically, based on the evaluation results, identify the aspects that the user needs to strengthen during the rehabilitation process, and adjust the specific exercise content, frequency, and duration in the training plan. The adjustment methods may include but are not limited to: Increasing the frequency of specific exercises: For example, if the user's pitch improvement is not obvious, it is recommended to increase pitch-related vocal exercises. Reduce the intensity of training: If the user experiences fatigue or discomfort during certain training, the intensity and duration of the training should be adjusted, and an adaptive training plan should be provided. In this way, the timeliness and accuracy of the rehabilitation effect can be improved.

[0143] Preferably, when generating the first rehabilitation training program and / or the second rehabilitation training program, the following steps are also included:

[0144] The first step is to obtain the target training days of the first user;

[0145] Specifically, it can be provided to the user in the form of an input box, radio buttons or a slider bar to ensure that the user's target training days are easy to input and adjust.

[0146] The second step is to determine the reasonable range of training days through the voice rehabilitation model based on the first user's medical history, symptom characteristics, diagnosis results and medical guidance standards;

[0147] The third step is to compare and analyze the target training days with the reasonable training days range determined by the voice rehabilitation model;

[0148] Step 4: When the target training days are within the reasonable training days range, a rehabilitation training plan of corresponding duration is generated based on the target training days;

[0149] Step 5: When the target number of training days is less than the minimum value of the reasonable training days range, a rehabilitation training plan is generated based on the minimum value of the reasonable training days range, and a prompt message is output to the user, indicating that the target number of training days is unreasonable. Specifically, the prompt message can be presented in the form of a pop-up window, notification, or information flow.

[0150] In this way, the voice rehabilitation model can be used to analyze the patient's health status and medical history, determine a reasonable range of training days, and help patients set a scientific and achievable rehabilitation goal, avoiding the risk of rehabilitation plan failure or overtraining due to blind pursuit of goals.

[0151] In addition, patients can be allowed to set their own training goals and adjust them according to the recommendations of the voice rehabilitation model, which enhances the patient's sense of participation and initiative in the rehabilitation process. When patients see that their goals are consistent with the model's recommendations, they will have more trust in the rehabilitation program and will be more actively involved in the training, which not only improves the effectiveness of the treatment, but also enhances the patient's satisfaction and confidence.

[0152] In addition, by comparing the target training days set by the patient with the reasonable range recommended by the model, on the one hand, a rehabilitation plan that is more in line with the patient's actual situation can be generated, ensuring that the patient can maintain an appropriate training intensity during the rehabilitation process, which helps to improve the overall rehabilitation effect and enable the patient to achieve the best rehabilitation state in a shorter period of time, reducing the risks that may arise during the rehabilitation process, especially the risk of voice damage caused by overtraining or improper training; on the other hand, it can ensure that the second rehabilitation training plan will not exceed the target training days initially set, ensuring the continuity and consistency of the rehabilitation process, so that patients can make steady progress within a scientific training framework.

[0153] Preferably, the personalized voice rehabilitation training method based on deep learning also includes the following steps:

[0154] The first step is to calculate the similarity index between the first rehabilitation training program of the first user and the first rehabilitation training programs of other users through a feature similarity analysis algorithm;

[0155] Specifically, the method includes the following steps: a first sub-step: performing feature extraction on first audio data involved in a first rehabilitation training program of a first user, wherein the extracted features include but are not limited to pitch (Fundamental Frequency), intensity (Intensity), timbre (Timbre), speech rate (Speech Rate), formant frequencies (Formant Frequencies), spectral features (Spectra l Features), etc.

[0156] Second sub-step: convert the extracted audio features into quantifiable feature vectors. Use cosine similarity, Euclidean distance, or Pearson correlation coefficient algorithm to compare and analyze the feature vectors, and calculate the similarity index between the feature vector of the first user and the feature vectors of other users.

[0157] In the second step, based on the similarity index, users whose similarity with the first user is higher than a preset threshold are screened out to form a similar user set.

[0158] Specifically, a preset similarity threshold can be set according to actual needs to screen other users whose voice rehabilitation training programs are highly similar to the first user. The threshold can be determined by methods such as cross-validation to optimize the accuracy of similar user screening.

[0159] In this way, on the one hand, the reference basis for personalized training programs can be enhanced, and the successful programs of others can be learned to improve the rehabilitation effect; on the other hand, it can facilitate the selection of rehabilitation partners, ensure that the rehabilitation partners have a high degree of relevance in voice characteristics, rehabilitation needs and training programs, promote more effective communication between partners, and enhance mutual support; establish a positive competitive atmosphere to further motivate users. In a partnership with common goals and similar rehabilitation paths, users are often motivated by the progress of their partners, which motivates them to train harder. This healthy competitive relationship can increase users' training enthusiasm and thus improve the overall rehabilitation effect.

[0160] In addition, it can also provide users with a platform for mutual supervision. During the rehabilitation process, users can observe and evaluate each other's progress, give timely feedback and suggestions, and thus form a positive supervision mechanism. This supervision not only enhances the user's sense of responsibility during the training process, but also reduces slackness and interruptions in training, making the rehabilitation process more continuous and stable.

[0161] In addition, through this partnership, users can share successful experiences and strategies for coping with challenges. This experience sharing is not only an encouragement to individuals, but also a positive promotion for the entire rehabilitation process. When users see successful cases of similar situations, they tend to increase their confidence and be more determined to implement rehabilitation training programs.

[0162] Furthermore, the personalized voice rehabilitation training method based on deep learning also includes the following steps:

[0163] The first step is to generate a corresponding recommended rehabilitation training plan according to the rehabilitation training plan of one of the users in the similar user set;

[0164] In the second step, the recommended rehabilitation training program is used as input, combined with the second audio data, and the first rehabilitation training program is adjusted through the voice rehabilitation model to obtain the second rehabilitation training program. In this way, it can ensure that the training progress of the current user is synchronized with the training progress of his rehabilitation partner. This can help maintain and enhance the partnership; on the other hand, it can improve the effectiveness of rehabilitation, allowing patients to psychologically feel the common progress with their rehabilitation partners on the road to rehabilitation, thereby enhancing the enthusiasm and persistence of treatment.

[0165] See also Figure 2Based on the same inventive concept as the personalized voice rehabilitation training method based on deep learning in the aforementioned embodiment, the embodiment of the present application provides a personalized voice rehabilitation training system based on deep learning, including:

[0166] A first acquisition module 201 is used to acquire a training sample set;

[0167] A first construction module 202 is used to train the deep learning model according to the training sample set to obtain a voice rehabilitation model;

[0168] The second acquisition module 203 is used to acquire the first audio data and the medical consultation information of the first user;

[0169] A first generating module 204, configured to generate a first rehabilitation training program based on the first audio data and the consultation information and through a voice rehabilitation model;

[0170] The third acquisition module 205 is used to acquire second audio data of the first user; the second audio data is audio data after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time;

[0171] The second generating module 206 is used to adjust the first rehabilitation training program based on the second audio data and through the voice rehabilitation model to generate a second rehabilitation training program.

[0172] In some embodiments, the training sample set includes historical audio data of multiple users, corresponding consultation information and related rehabilitation result data; the consultation information is recorded by a professional physician, and the consultation information includes the user's medical history, symptom description, diagnosis results, and recommended treatment plan; the rehabilitation result data includes follow-up audio data of the user after completing the recommended treatment plan and the rehabilitation result data corresponding to the audio data;

[0173] The first construction module 202 is specifically used for: preprocessing the audio data of each user; labeling the preprocessed audio features and the corresponding medical consultation information and rehabilitation result data to form one-to-one corresponding training samples; inputting multiple training samples into the deep learning model for training to obtain a voice rehabilitation model.

[0174] In some embodiments, the first generation module 204 is specifically used to: pre-process the first audio data to extract audio features including volume, pitch, resonance peaks, duration, and speaking speed; perform semantic analysis on the symptom description, medical history information, and diagnosis results in the medical consultation information to extract corresponding feature labels; associate the audio features with the feature labels of the medical consultation information and input them as input data into the voice rehabilitation model; the voice rehabilitation model analyzes the user's voice status based on the input audio features and the feature labels of the medical consultation information to obtain analysis results; and generate a first rehabilitation training plan based on the analysis results.

[0175] In some embodiments, the second generation module 206 is specifically used to: convert the first audio data and the second audio data into a feature vector that can be used for model analysis through a feature extraction algorithm; compare the feature vector of the second audio data with the feature vector of the first audio data, and calculate the difference index between the two; obtain the rehabilitation progress according to the difference index; adjust the training parameters of the first rehabilitation training plan according to the rehabilitation progress, and obtain the second rehabilitation training plan.

[0176] In some embodiments, the first generation module 204 and / or the second generation module 206 are also specifically used to: obtain the target training days for the first user; determine a reasonable range of training days through a voice rehabilitation model based on the first user's medical history, symptom characteristics, diagnosis results and medical guidance standards; compare and analyze the target training days with the reasonable range of training days determined by the voice rehabilitation model; when the target training days are within the reasonable range of training days, generate a rehabilitation training plan of corresponding duration based on the target training days; when the target training days are less than the minimum value of the reasonable range of training days, generate a rehabilitation training plan based on the minimum value of the reasonable range of training days, and output a prompt message to the user, pointing out the unreasonableness of the target training days.

[0177] In some embodiments, the personalized voice rehabilitation training system based on deep learning also includes a first screening module, which is used to calculate the similarity index between the first rehabilitation training plan of the first user and the first rehabilitation training plans of other users through a feature similarity analysis algorithm; based on the similarity index, users whose similarity with the first user is higher than a preset threshold are screened out to form a set of similar users.

[0178] In some embodiments, the second generation module 206 is specifically used to: generate a corresponding recommended rehabilitation training plan based on the rehabilitation training plan of one of the users in the similar user set; take the recommended rehabilitation training plan as input, combine it with the second audio data, adjust the first rehabilitation training plan through the voice rehabilitation model, and obtain the second rehabilitation training plan.

[0179] It is understandable that the modules and references recorded in the personalized voice rehabilitation training system based on deep learning Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the personalized voice rehabilitation training system based on deep learning and the modules contained therein, and will not be repeated here.

[0180] See also Figure 3 , based on the inventive concept of a personalized voice rehabilitation training method based on deep learning in the aforementioned embodiment, an embodiment of the present application provides a terminal device. The terminal device may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), etc., and fixed terminals such as digital TVs, desktop computers, etc. The terminal device includes a processing device 301 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a ROM 302 (read-only memory) or a program loaded from a storage device 308 to a RAM 303 (random access memory). In RAM 303, various programs and data required for the operation of the terminal device are also stored. The processing device 301, ROM 302, and RAM 303 are connected to each other via a bus 304. The input / output interface (i.e., an I / O interface 305) is also connected to the bus 304.

[0181] Typically, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the terminal device to communicate with other devices wirelessly or by wire to exchange data.

[0182] In particular, according to some embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present application are executed.

[0183] It should be noted that the computer-readable medium recorded in some embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In some embodiments of the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0184] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0185] The computer-readable medium may be included in the terminal device, or may exist independently without being installed in the terminal device. The computer-readable medium carries one or more programs, and when the one or more programs are executed by the terminal device, the method steps of any of the above technical solutions may be implemented.

[0186] The computer program code for performing the operation of some embodiments of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0187] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0188] The modules described in some embodiments of the present application may be implemented in software or hardware. The modules described may also be arranged in the processing device 301 .

[0189] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0190] Some embodiments of the present application also provide a computer program product, including a computer program, which implements any of the above-mentioned personalized voice rehabilitation training methods based on deep learning when executed by the processing device 301.

[0191] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0192] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A personalized voice rehabilitation training method based on deep learning, characterized in that: The following steps are involved: Get a training sample set; Training the deep learning model according to the training sample set to obtain a voice rehabilitation model; Acquiring first audio data and medical consultation information of a first user; Based on the first audio data and the consultation information, and using the voice rehabilitation model, a first rehabilitation training program is generated; Acquire second audio data of the first user; The second audio data is audio data after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time; Based on the second audio data, the first rehabilitation training program is adjusted through the voice rehabilitation model to generate a second rehabilitation training program.

2. The personalized voice rehabilitation training method based on deep learning according to claim 1, characterized in that: The training sample set includes historical audio data of multiple users, corresponding medical consultation information and related rehabilitation result data; the medical consultation information is recorded by professional physicians, and the medical consultation information includes the user's medical history, symptom description, diagnosis results, and recommended treatment plan; the rehabilitation result data includes the follow-up audio data of the user after completing the recommended treatment plan and the rehabilitation result data corresponding to the audio data; The step of training the deep learning model according to the training sample set to obtain the voice rehabilitation model comprises: Preprocessing the historical audio data of each user to extract audio features including volume, pitch, formant, duration, and speech speed; Labeling the audio features with the corresponding medical consultation information and rehabilitation result data to form one-to-one corresponding training samples; Multiple training samples are input into the deep learning model for training to obtain a voice rehabilitation model.

3. The personalized voice rehabilitation training method based on deep learning according to claim 1 or 2, characterized in that: The generation of the first rehabilitation training program and / or the second rehabilitation training program also includes the following steps: Obtaining the target number of training days for the first user; Determining a reasonable range of training days through the voice rehabilitation model based on the medical history, symptom characteristics, diagnosis results and medical guidance standards of the first user; Comparing and analyzing the target training days with the reasonable training days range determined by the voice rehabilitation model; When the target number of training days is within the range of the reasonable number of training days, generating a rehabilitation training program of corresponding duration based on the target number of training days; When the target training days are less than the minimum value of the reasonable training days range, a rehabilitation training plan is generated based on the minimum value of the reasonable training days range, and prompt information is output to the user.

4. The personalized voice rehabilitation training method based on deep learning according to claim 1, characterized in that: The step of adjusting the first rehabilitation training program based on the second audio data and using the voice rehabilitation model to generate a second rehabilitation training program includes: Converting the first audio data and the second audio data into feature vectors that can be used for model analysis through a feature extraction algorithm; Comparing the feature vector of the second audio data with the feature vector of the first audio data, and calculating a difference index between the two; According to the difference index, the rehabilitation progress is obtained; The training parameters of the first rehabilitation training program are adjusted according to the rehabilitation progress to obtain a second rehabilitation training program.

5. The personalized voice rehabilitation training method based on deep learning according to claim 1, characterized in that: The step of generating a first rehabilitation training program based on the first audio data and the consultation information and through the voice rehabilitation model comprises: Preprocessing the first audio data to extract audio features including volume, pitch, formant, duration, and speech speed; Perform semantic analysis on the symptom description, medical history information, and diagnosis results in the medical consultation information to extract corresponding feature tags; Associating the audio features with the feature labels of the medical consultation information and inputting the results as input data into the voice rehabilitation model; The voice rehabilitation model analyzes the user's voice status based on the input audio features and the feature labels of the medical consultation information to obtain an analysis result; The first rehabilitation training program is generated according to the analysis result.

6. The personalized voice rehabilitation training method based on deep learning according to claim 1, characterized in that: The following steps are also included: Calculating the similarity index between the first rehabilitation training program of the first user and the first rehabilitation training programs of other users by using a feature similarity analysis algorithm; Based on the similarity index, users whose similarity with the first user is higher than a preset threshold are screened out to form a similar user set.

7. A personalized voice rehabilitation training system based on deep learning, characterized in that: include: A first acquisition module is used to acquire a training sample set; A first construction module is used to train the deep learning model according to the training sample set to obtain a voice rehabilitation model; A second acquisition module is used to acquire the first audio data and medical consultation information of the first user; A first generating module, configured to generate a first rehabilitation training program based on the first audio data and the consultation information and through the voice rehabilitation model; A third acquisition module, used to acquire second audio data of the first user; The second audio data is audio data after the first user has undergone rehabilitation training according to the first rehabilitation training program and reached a first preset training time; The second generating module is used to adjust the first rehabilitation training program based on the second audio data and through the voice rehabilitation model to generate a second rehabilitation training program.

8. A terminal device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images