Risk assessment device based on artificial intelligence, computer equipment and storage medium

Abstract

The invention belongs to the field of artificial intelligence and the field of digital medical treatment, and relates to an artificial intelligence-based risk assessment device, computer equipment and a storage medium, and the device comprises the steps: obtaining target voice data of a target user through the device; performing audio conversion on the target voice data to obtain a spectrogram; converting the spectrogram to obtain an embedded vector; screening out a target risk assessment model from all pre-trained risk assessment models; and evaluating the embedded vector based on the target risk evaluation model to generate a risk evaluation result of the target user. The method can be applied to risk assessment scenes in the field of digital medical treatment. In addition, the risk assessment result can be stored in the block chain. According to the method and the device, risk assessment is performed on the voice data based on the target risk assessment model, so that the risk assessment efficiency of judging whether the user has the risk of suffering from the Alzheimer's disease or not is effectively improved, and the accuracy of the risk assessment result is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and digital medicine, and in particular to an artificial intelligence-based risk assessment device, computer equipment, and storage medium. Background Art

[0002] Alzheimer's disease is a progressive neurodegenerative disease that causes problems with memory, thinking and behavior. It progresses rapidly, has a high prevalence rate and high long-term care costs. As a degenerative disease of the central nervous system, it has a slow and insidious onset, and patients and their families cannot tell when the disease started. Its clinical manifestations include decreased cognitive ability, memory loss, and difficulty concentrating, which seriously affect daily life and are accompanied by brain atrophy.

[0003] In the prior art, electroencephalograms (EEGs) are commonly used in medicine as a basis for risk analysis of Alzheimer's disease. EEGs have high temporal resolution and are relatively cheap. Doctors obtain EEGs and then give risk assessment results based on their own medical experience. This risk assessment method requires a lot of clinical experience from doctors, and has slow processing efficiency and low accuracy. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a risk assessment device, computer equipment and storage medium based on artificial intelligence to solve the problem that in the prior art, electroencephalograms are usually used in medicine as the basis for risk analysis of Alzheimer's disease. Doctors obtain the electroencephalograms and then give risk assessment results based on their own medical experience. This risk assessment method requires a lot of clinical accumulation from doctors, and has technical problems such as slow processing efficiency and low accuracy.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a risk assessment device based on artificial intelligence, which adopts the following technical solutions:

[0006] A first acquisition module, used to acquire target voice data of a target user;

[0007] A first processing module, used for performing audio conversion processing on the target speech data to obtain a corresponding spectrogram;

[0008] A second processing module, configured to perform conversion processing on the spectrum graph based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrum graph;

[0009] A calling module, used to call a plurality of pre-trained risk assessment models; wherein the plurality of risk assessment models are models generated by training a plurality of pre-trained models having different pre-trained weights respectively using a sample data set consisting of pre-collected speech data related to Alzheimer's disease;

[0010] A screening module, used to screen out a target risk assessment model from all the risk assessment models based on preset screening rules;

[0011] An evaluation module is used to evaluate the embedding vector based on the target risk evaluation model to generate a risk evaluation result corresponding to the target user.

[0012] Furthermore, the risk assessment device based on artificial intelligence also includes:

[0013] A second acquisition module is used to acquire pre-collected speech data related to Alzheimer's disease;

[0014] an enhancement module, configured to perform an acoustic enhancement operation on the voice data to obtain corresponding first voice data;

[0015] A labeling module, used for performing data labeling processing on the first voice data to obtain corresponding second voice data;

[0016] A loading module, used to load a specified pre-trained model corresponding to the specified pre-trained weights; wherein the specified pre-trained model is any one of all the pre-trained models;

[0017] A conversion module, used for converting the second speech data to obtain a sample data set;

[0018] A partitioning module, used to divide the sample data set into a training set, a validation set and a test set;

[0019] The training module is used to train the specified pre-trained model using the training set, and to fine-tune the specified pre-trained model using a preset optimization algorithm to obtain a trained first risk assessment model.

[0020] Furthermore, the enhancement module includes:

[0021] A first processing submodule, configured to perform time distortion processing on the voice data to obtain corresponding third voice data;

[0022] A second processing submodule is used to perform frequency domain masking processing on the voice data to obtain corresponding fourth voice data;

[0023] A third processing submodule performs time domain masking processing on the voice data to obtain corresponding fifth voice data;

[0024] A construction submodule is used to construct the first voice data based on the third voice data, the fourth voice data and the fifth voice data.

[0025] Furthermore, the screening module includes:

[0026] A verification submodule, used to perform performance verification processing on each of the risk assessment models using the verification set to obtain a performance indicator value of each of the risk assessment models corresponding to a preset performance indicator;

[0027] A first calculation submodule, used to calculate the performance evaluation score of each risk evaluation model based on the performance indicator value of each risk evaluation model;

[0028] A screening submodule, used to screen out a second risk assessment model with the highest performance assessment score from all the risk assessment models;

[0029] A first testing submodule, configured to test the second risk assessment model based on the test set to determine whether the second risk assessment model meets preset performance expectations;

[0030] The first determining submodule is configured to, if yes, use the second risk assessment model as the target risk assessment model.

[0031] Furthermore, the performance index includes at least a first performance index, a second performance index and a third performance index; and the first calculation submodule includes:

[0032] A first acquisition unit is used to acquire first performance indicator data corresponding to the first performance indicator, second performance indicator data corresponding to the second performance indicator, and third performance indicator data corresponding to the third performance indicator from the performance indicator data of a specified risk assessment model; wherein the specified risk assessment model is any one of all the risk assessment models;

[0033] A second acquisition unit, configured to acquire a first weight corresponding to the first performance indicator, a second weight corresponding to the second performance indicator, and a third weight corresponding to the third performance indicator;

[0034] A calling unit, used to call a preset calculation formula;

[0035] a calculation unit, configured to calculate and process the first performance indicator data, the second performance indicator data, the third performance indicator data, the first weight, the second weight, and the third weight based on the calculation formula to obtain a corresponding calculation result;

[0036] A determination unit is used to use the calculation result as a performance evaluation score of the specified risk assessment model.

[0037] Furthermore, the screening module further includes:

[0038] An acquisition submodule, configured to acquire a preset model optimization strategy if the second risk assessment model does not meet the performance expectation;

[0039] An optimization submodule, used to optimize the second risk assessment model based on the model optimization strategy to obtain a corresponding optimization model;

[0040] A second testing submodule is used to test the optimization model using the test set to obtain a corresponding test performance indicator value;

[0041] A second calculation submodule, configured to calculate a specified performance evaluation score of the optimization model based on the test performance indicator;

[0042] A judgment submodule, used to judge whether the specified performance evaluation score is greater than a preset score threshold;

[0043] The second determining submodule is used to, if yes, use the optimization model as the target risk assessment model.

[0044] Furthermore, the second processing module includes:

[0045] The calling submodule is used to call the pre-trained convolutional neural network;

[0046] An input submodule, used for inputting the spectrum graph into the convolutional neural network, processing the spectrum graph by the convolutional neural network, and obtaining an output result corresponding to the spectrum graph;

[0047] The third determination submodule is used to use the output result as the embedding vector.

[0048] Furthermore, the risk assessment device based on artificial intelligence also includes:

[0049] A determination module, used to determine a target storage strategy corresponding to the target risk assessment model;

[0050] A storage module performs storage processing on the target risk assessment model based on the target storage strategy.

[0051] In order to solve the above technical problems, an embodiment of the present application also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the functions of each module in the above-mentioned artificial intelligence-based risk assessment device are implemented.

[0052] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the functions of each module in the above-mentioned artificial intelligence-based risk assessment device are realized.

[0053] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0054] In the present application, an artificial intelligence-based risk assessment device first obtains the target voice data of the target user; then performs audio conversion processing on the target voice data to obtain a corresponding spectrogram; then converts the spectrogram based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrogram; subsequently calls multiple pre-trained risk assessment models; further filters out a target risk assessment model from all the risk assessment models based on preset screening rules; finally, evaluates the embedding vector based on the target risk assessment model to generate a risk assessment result corresponding to the target user. The present application obtains a corresponding embedding vector by performing audio conversion and conversion processing on the acquired target voice data of the target user, and then screens out a target risk assessment model from all pre-trained risk assessment models based on screening rules, and subsequently uses the target risk assessment model to perform evaluation processing on the embedding vector to generate a risk assessment result corresponding to the target user. Different from the existing method in which doctors give risk assessment results for the acquired electroencephalograms based on their own medical experience, the present application performs risk assessment processing on the target voice data of the target user based on the use of a target risk assessment model, which can effectively improve the efficiency of risk assessment for determining whether a user has a risk of suffering from Alzheimer's disease, and improve the accuracy of the generated risk assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0057] Figure 2 is a schematic structural diagram of an embodiment of an artificial intelligence-based risk assessment device according to the present application;

[0058] Figure 3 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0060] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0061] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0062] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0063] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0064] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0065] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0066] It should be noted that the artificial intelligence-based risk assessment device provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based risk assessment device is generally set in the server / terminal device.

[0067] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0068] Continue to refer Figure 2 , showing a structural diagram of an embodiment of an artificial intelligence-based risk assessment device according to the present application. According to different needs, the order of the modules in the structural diagram can be changed, and some modules can be omitted. The artificial intelligence-based risk assessment device provided in the embodiment of the present application can be applied to any risk assessment scenario that requires voice data related to Alzheimer's disease, and the artificial intelligence-based risk assessment device can be applied to products in these scenarios. The artificial intelligence-based risk assessment device 200 includes: a first acquisition module 201, a first processing module 202, a second processing module 203, a calling module 204, a screening module 205 and an evaluation module 206. Among them:

[0069] The first acquisition module 201 is used to acquire target voice data of a target user.

[0070] In this embodiment, the electronic device (e.g. Figure 1The target voice data can be obtained by a wired connection or a wireless connection. It should be noted that the wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi DeBand) connection, and other wireless connection methods known now or to be developed in the future. The present application can be applied to the field of digital medicine, and can be specifically used to assess whether a user has a medical business scenario at risk of suffering from Alzheimer's disease. The target user may be a potential user to be detected who may have a risk of suffering from Alzheimer's disease, and the target voice data may be voice data emitted by the target user collected by a voice acquisition device during a cognitive function assessment test. Among them, after completing the acquisition of the target voice data, the voice data can also be noise eliminated, and the volume of all voice clips can be standardized to control changes caused by recording conditions (such as microphone placement).

[0071] The first processing module 202 is used to perform audio conversion processing on the target speech data to obtain a corresponding spectrogram.

[0072] In this embodiment, after obtaining the target voice data, it is also necessary to convert the target voice data into a format that can be processed by the risk assessment model. Since the risk assessment model is trained based on the pre-trained BERT model, it is necessary to convert the target voice data into a spectrogram.

[0073] The second processing module 203 is used to perform conversion processing on the spectrum graph based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrum graph.

[0074] In this embodiment, after obtaining the spectrogram, it is also necessary to perform adaptive adjustments on the spectrogram related to the pre-trained BERT model. Since the BERT model directly processes word embeddings, the spectrogram needs to be further converted into a corresponding embedding vector. Among them, the specific implementation process of converting the spectrogram based on the preset conversion strategy to obtain the embedding vector corresponding to the spectrogram will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0075] The calling module 204 is used to call multiple pre-trained risk assessment models; wherein the multiple risk assessment models are models generated by training multiple pre-trained models with different pre-training weights respectively using a sample data set composed of pre-collected speech data related to Alzheimer's disease.

[0076] In this embodiment, according to the sample data set consisting of the pre-collected speech data related to Alzheimer's disease, the specific implementation process of training and generating the model for multiple pre-trained models with different pre-trained weights is described in further detail in the subsequent specific embodiments of this application, and will not be elaborated on here. Among them, there is no specific limitation on the number of risk assessment models, which can be set according to actual training needs. Exemplarily, if there are K pre-trained models, the K pre-trained models will be respectively set with different pre-trained weights, and the number of pre-trained weights is also K, and then the same training process is performed on the multiple pre-trained models by using the sample data set consisting of the pre-collected speech data related to Alzheimer's disease, so as to obtain multiple trained risk assessment models.

[0077] The screening module 205 is used to screen out a target risk assessment model from all the risk assessment models based on a preset screening rule.

[0078] In this embodiment, the specific implementation process of screening out the target risk assessment model from all the risk assessment models based on the preset screening rules will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0079] The evaluation module 206 is used to evaluate the embedding vector based on the target risk evaluation model to generate a risk evaluation result corresponding to the target user.

[0080] In this embodiment, the embedding vector is input into the target risk assessment model, the embedding vector is evaluated and processed by the target risk assessment model, and then a risk assessment result corresponding to the target user is output, and the risk assessment result is used to assess whether the user has a risk of suffering from Alzheimer's disease.

[0081] In the present application, an artificial intelligence-based risk assessment device obtains a corresponding embedding vector by performing audio conversion and conversion processing on the acquired target voice data of the target user, and then screens out a target risk assessment model from all pre-trained risk assessment models based on screening rules, and subsequently uses the target risk assessment model to perform evaluation processing on the embedding vector to generate a risk assessment result corresponding to the target user. Different from the existing way in which doctors give risk assessment results for the acquired electroencephalograms based on their own medical experience, the present application is based on the method of performing risk assessment processing on the target voice data of the target user using a target risk assessment model, which can effectively improve the efficiency of risk assessment for determining whether a user has a risk of suffering from Alzheimer's disease, and improve the accuracy of the generated risk assessment results.

[0082] In some optional implementations, the above-mentioned artificial intelligence-based risk assessment device further includes:

[0083] The second acquisition module is used to acquire pre-collected speech data related to Alzheimer's disease.

[0084] In this embodiment, the above-mentioned voice data related to Alzheimer's disease may at least include voice data of Alzheimer's disease patients and voice data of cognitively normal people. The voice data of Alzheimer's disease patients and the voice data of cognitively normal people can be collected and recorded in a cognitive function assessment test. In addition, after obtaining the voice data related to Alzheimer's disease, the voice data can be further cleaned, such as removing noise, to ensure the accuracy of the voice data.

[0085] The enhancement module is used to perform an acoustic enhancement operation on the voice data to obtain corresponding first voice data.

[0086] In this embodiment, the above acoustic enhancement operation may specifically adopt the SpecAugment enhancement operation. SpecAugment is an enhancement method that operates on the log spectrum of the input audio instead of the original audio. The enhancement strategies include time warping, frequency domain masking and time domain masking.

[0087] The annotation module is used to perform data annotation processing on the first voice data to obtain corresponding second voice data.

[0088] In this embodiment, the first voice data is annotated to clarify which voices in the first voice data belong to the Alzheimer's disease category and which voices belong to the non-Alzheimer's disease category. The data annotation process can be performed by machine annotation or manual annotation.

[0089] A loading module is used to load a specified pre-trained model corresponding to the specified pre-trained weights; wherein the specified pre-trained model is any one of all the pre-trained models.

[0090] In this embodiment, the above-mentioned pre-trained model can specifically adopt a BERT model. Since the BERT model has usually been pre-trained on a large-scale corpus, it can be directly used for downstream tasks. Among them, multiple BERT models with different pre-trained weights are pre-adopted to train and generate corresponding multiple risk assessment models.

[0091] The conversion module is used to convert the second speech data to obtain a sample data set.

[0092] In this embodiment, the conversion process includes first converting the second speech data into a corresponding designated spectrogram, then converting the designated spectrogram into a corresponding designated embedding vector, and using the designated embedding vector as the sample data set.

[0093] The partitioning module is used to partition the sample data set into a training set, a validation set and a test set.

[0094] In this embodiment, the sample data set can be divided into a training set, a validation set and a test set according to a preset division ratio. The value of the division ratio is not specifically limited and can be set according to actual business usage requirements, for example, it can be set to 6:2:2.

[0095] The training module is used to train the specified pre-trained model using the training set, and to fine-tune the specified pre-trained model using a preset optimization algorithm to obtain a trained first risk assessment model.

[0096] In this embodiment, the specified pre-trained model can be trained using the training set based on a supervised learning method, specifically by minimizing the loss function (such as cross entropy loss), optimizing the parameters of the model, and using optimization algorithms such as gradient descent to fine-tune the model, and by adding a classification layer on top of the specified pre-trained model, usually a fully connected layer, for mapping the embedded representation output by the specified pre-trained model to the final classification result, thereby obtaining a trained first risk assessment model. After completing the training of the risk assessment model, the risk assessment model can be used to conduct risk assessment on the user's voice data, by inputting the voice data after patch embedding into the risk assessment model, and classifying it through class token to determine whether the user belongs to the Alzheimer's disease category, and outputting the risk assessment result corresponding to whether the user has the risk of suffering from Alzheimer's disease.

[0097] The present application obtains a sample data set by performing acoustic enhancement operations, data annotation processing and conversion processing on pre-collected speech data related to Alzheimer's disease, and then uses the training set in the sample data set to train the loaded specified pre-trained model corresponding to the specified pre-trained weights. This can quickly and conveniently build a trained risk assessment model, so that the risk assessment model can be used subsequently to accurately analyze the speech data input by the user to identify the risk assessment result of whether the user suffers from Alzheimer's disease.

[0098] In some optional implementations, the enhancement module includes:

[0099] The first processing submodule is used to perform time distortion processing on the voice data to obtain corresponding third voice data.

[0100] In this embodiment, the time warping process refers to performing time warping on the audio signal to simulate the change of speech speed.

[0101] The second processing submodule performs frequency domain masking processing on the voice data to obtain corresponding fourth voice data.

[0102] In this embodiment, frequency domain masking refers to randomly selecting a portion on the frequency axis of the spectrogram to mask, that is, setting the values ​​of these frequency segments to 0 to simulate noise or spectrum loss. The speech data may be first converted to obtain the corresponding spectrogram.

[0103] The third processing submodule performs time domain masking processing on the voice data to obtain corresponding fifth voice data.

[0104] In this embodiment, the time domain masking process refers to randomly selecting a portion on the time axis of the spectrogram for masking, that is, setting the values ​​of these time periods to 0 to simulate the absence or noise of the audio segment.

[0105] A construction submodule is used to construct the first voice data based on the third voice data, the fourth voice data and the fifth voice data.

[0106] In this embodiment, the third voice data, the fourth voice data and the fifth voice data are integrated to obtain integrated voice data, and the integrated voice data is used as the first voice data, wherein the first voice data at least includes the third voice data, the fourth voice data and the fifth voice data.

[0107] The present application performs time warping processing on the voice data to obtain corresponding third voice data; and performs frequency domain masking processing on the voice data to obtain corresponding fourth voice data; and performs time domain masking processing on the voice data to obtain corresponding fifth voice data; and subsequently constructs the first voice data based on the third voice data, the fourth voice data and the fifth voice data. The present application performs time warping processing, frequency domain masking processing and time domain masking processing on the voice data to complete the processing of extended training data for the voice data by performing acoustic enhancement operations on the voice data, thereby ensuring the data richness and diversity of the obtained first voice data.

[0108] In some optional implementations of this embodiment, the screening module 205 includes:

[0109] The verification submodule is used to perform performance verification processing on each of the risk assessment models using the verification set to obtain performance indicator values ​​of each of the risk assessment models corresponding to preset performance indicators.

[0110] In this embodiment, the validation set is input into each of the risk assessment models, so that each of the risk assessment models performs risk assessment on the validation set, and then the performance of each of the risk assessment models is verified according to the output results of each of the risk assessment models, that is, each of the risk assessment models is evaluated to obtain the performance index values ​​of each of the risk assessment models. The above performance index values ​​may include specific values ​​corresponding to the indicators such as accuracy, recall rate, F1 value, etc. calculated for each risk assessment model to evaluate the performance of the model.

[0111] The first calculation submodule is used to calculate the performance evaluation score of each risk evaluation model based on the performance indicator value of each risk evaluation model.

[0112] In this embodiment, the specific implementation process of calculating the performance evaluation score of each risk assessment model based on the performance indicator value of each risk assessment model will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated here.

[0113] The screening submodule is used to screen out the second risk assessment model with the highest performance assessment score from all the risk assessment models.

[0114] In this embodiment, the performance evaluation scores of the risk evaluation models can be numerically compared, and then the second risk evaluation model with the highest performance evaluation score can be screened out from all the risk evaluation models according to the obtained numerical comparison results.

[0115] The first testing submodule is used to test the second risk assessment model based on the test set to determine whether the second risk assessment model meets preset performance expectations.

[0116] In this embodiment, the second risk assessment model is tested using the test set to obtain a specific test performance indicator value corresponding to the second risk assessment model, and then the specific performance evaluation score of the second risk assessment model is calculated based on the specific test performance indicator value, and then the specific performance evaluation score is compared with the preset score threshold. If the specific performance evaluation score is greater than the above score threshold, it is determined that the second risk assessment model meets the performance expectations, and if the specific performance evaluation score is less than the above score threshold, it is determined that the second risk assessment model does not meet the performance expectations. Among them, there is no specific limitation on the value of the above score threshold, which can be set according to the actual performance test requirements.

[0117] The first determining submodule is configured to, if yes, use the second risk assessment model as the target risk assessment model.

[0118] The present application performs performance verification processing on each of the risk assessment models by using the verification set to obtain the performance indicator value of each of the risk assessment models; then calculates the performance evaluation score of each of the risk assessment models based on the performance indicator value of each of the risk assessment models; then screens out the second risk assessment model with the highest performance evaluation score from all the risk assessment models; subsequently, tests the second risk assessment model based on the test set to determine whether the second risk assessment model meets the preset performance expectations; if so, uses the second risk assessment model as the target risk assessment model. The present application performs performance verification processing on each of the risk assessment models using a verification set to screen out the second risk assessment model with the highest performance assessment score from all the risk assessment models, and intelligently tests the second risk assessment model based on the test set. When it is determined that the second risk assessment model meets the preset performance expectations, the second risk assessment model is used as the target risk assessment model to accurately use the second risk assessment model with the best performance among all risk assessment models as the final target risk assessment model responsible for performing risk assessment processing on the user's voice data, thereby effectively ensuring the accuracy of the target risk assessment model, facilitating the subsequent use of the target risk assessment model to evaluate and process the target voice data of the target user, and effectively improving the accuracy of the risk assessment results generated for the target user.

[0119] In some optional implementations, the performance indicator includes at least a first performance indicator, a second performance indicator, and a third performance indicator; and the first calculation submodule includes:

[0120] A first acquisition unit is used to acquire first performance indicator data corresponding to the first performance indicator, second performance indicator data corresponding to the second performance indicator, and third performance indicator data corresponding to the third performance indicator from the performance indicator data of a specified risk assessment model; wherein the specified risk assessment model is any one of all the risk assessment models.

[0121] In this embodiment, the specified risk assessment model is any one of all the risk assessment models. The first performance indicator specifically refers to the accuracy, the second performance indicator specifically refers to the recall, and the third performance indicator specifically refers to the F1 value.

[0122] The second acquisition unit is used to acquire a first weight corresponding to the first performance indicator, a second weight corresponding to the second performance indicator, and a third weight corresponding to the third performance indicator.

[0123] In this embodiment, there is no specific limitation on the numerical selection of the first weight, the second weight and the third weight, which can be set according to actual performance statistical requirements, and the sum of the first weight, the second weight and the third weight is 1.

[0124] The calling unit is used to call the preset calculation formula.

[0125] In this embodiment, the above calculation formula is specifically: S=h*x+i*y+j*z, S is the performance evaluation score, x is the first performance indicator, h is the first weight, y is the second performance indicator, i is the second weight, z is the third performance indicator, and j is the third weight.

[0126] A calculation unit is used to calculate and process the first performance indicator data, the second performance indicator data, the third performance indicator data, the first weight, the second weight and the third weight based on the calculation formula to obtain a corresponding calculation result.

[0127] In this embodiment, the calculation result can be obtained by substituting the first performance indicator data, the second performance indicator data, the third performance indicator data, the first weight, the second weight and the third weight into the corresponding positions in the calculation formula.

[0128] A determination unit is used to use the calculation result as a performance evaluation score of the specified risk assessment model.

[0129] The present application obtains a first weight corresponding to the first performance indicator, a second weight corresponding to the second performance indicator, and a third weight corresponding to the third performance indicator, and then calculates the first performance indicator data, the second performance indicator data, and the third performance indicator data of the specified risk assessment model based on the use of a calculation formula, so as to quickly and accurately generate a performance evaluation score for the specified risk assessment model, thereby improving the efficiency of generating the performance evaluation score and ensuring the accuracy of the obtained performance evaluation score.

[0130] In some optional implementations of this embodiment, the screening module 205 further includes:

[0131] The acquisition submodule is used to obtain a preset model optimization strategy if the second risk assessment model does not meet the performance expectations.

[0132] In this embodiment, the model optimization strategy includes processing strategies such as adjusting model parameters / training parameters, modifying the model architecture, replacing the classification layer structure, or adding data enhancement strategies.

[0133] The optimization submodule is used to optimize the second risk assessment model based on the model optimization strategy to obtain a corresponding optimized model.

[0134] In this embodiment, the second risk assessment model may be optimized by adopting any one or more processing strategies in the model optimization strategy to obtain a corresponding optimization model.

[0135] The second testing submodule is used to use the test set to test the optimization model to obtain corresponding test performance indicator values.

[0136] In this embodiment, the test set is input into the optimization model so as to perform risk assessment on the test set through the optimization model, and the test performance indicator value corresponding to the optimization model is statistically obtained.

[0137] The second calculation submodule is used to calculate the specified performance evaluation score of the optimization model based on the test performance indicator.

[0138] In this embodiment, the calculation process of calculating the specified performance evaluation score of the optimization model based on the test performance indicator can refer to the calculation process of calculating the performance evaluation score of each risk assessment model based on the performance indicator value of each risk assessment model, and no further details will be given here.

[0139] The judgment submodule is used to judge whether the specified performance evaluation score is greater than a preset score threshold.

[0140] In this embodiment, there is no specific limitation on the value of the above-mentioned score threshold, which can be set according to actual performance test requirements.

[0141] The second determining submodule is used to, if yes, use the optimization model as the target risk assessment model.

[0142] In this embodiment, if the specified performance evaluation score is greater than the above-mentioned score threshold, it is determined that the optimization model meets the performance expectations, and the optimization model is used as the target risk evaluation model.

[0143] If the present application detects that the second risk assessment model does not meet the performance expectations, a preset model optimization strategy is obtained; then the second risk assessment model is optimized based on the model optimization strategy to obtain a corresponding optimization model; then the optimization model is tested using the test set to obtain a corresponding test performance indicator value; subsequently, a specified performance evaluation score of the optimization model is calculated based on the test performance indicator, and it is determined whether the specified performance evaluation score is greater than a preset score threshold; if so, the optimization model is used as the target risk assessment model. When the present application detects that the second risk assessment model does not meet performance expectations, it will also intelligently optimize the second risk assessment model using a model optimization strategy to obtain a corresponding optimization model, and then use a test set to calculate a specified performance evaluation score for the optimization model. When it is determined that the specified performance evaluation score is greater than a preset score threshold, that is, when it is determined that the optimization model has optimal performance and that the optimization model meets performance expectations, the optimization model is used as the final target risk assessment model responsible for performing risk assessment processing on the user's voice data, thereby effectively ensuring the accuracy of the target risk assessment model, facilitating the subsequent use of the target risk assessment model to evaluate and process the target voice data of the target user, and can effectively improve the accuracy of the risk assessment results generated for the target user.

[0144] In some optional implementations of this embodiment, the second processing module 203 includes:

[0145] The calling submodule is used to call the pre-trained convolutional neural network.

[0146] In this embodiment, a convolutional neural network (CNN) is a deep learning model commonly used in the fields of image recognition, speech recognition, etc. The core idea of ​​CNN is to extract features of data such as images through convolution operations, thereby achieving tasks such as data classification and recognition.

[0147] An input submodule is used to input the spectrum graph into the convolutional neural network, process the spectrum graph through the convolutional neural network, and obtain an output result corresponding to the spectrum graph.

[0148] In this embodiment, based on the structure of a convolutional neural network, the spectrogram is taken as input and an embedding vector corresponding to the spectrogram is output.

[0149] The third determination submodule is used to use the output result as the embedding vector.

[0150] In this embodiment, the above conversion strategy can also use a method similar to that in Vision Transformer (Vision Transformer), which divides the spectrogram into multiple small image patches, and then flattens and linearly maps each patch to an embedding vector.

[0151] The present application calls a pre-trained convolutional neural network; then inputs the spectrum graph into the convolutional neural network, processes the spectrum graph through the convolutional neural network, and obtains an output result corresponding to the spectrum graph; and subsequently uses the output result as the embedding vector. The present application processes the spectrum graph based on a pre-trained convolutional neural network, and can quickly and accurately generate an embedding vector corresponding to the spectrum graph, thereby improving the generation efficiency of the embedding vector and ensuring the accuracy of the generated embedding vector.

[0152] In some optional implementations of this embodiment, the above-mentioned risk assessment device based on artificial intelligence further includes:

[0153] A determination module is used to determine a target storage strategy corresponding to the target risk assessment model.

[0154] In this embodiment, there is no specific limitation on the selection of the above target storage strategy, which can be determined according to the actual storage business needs. Exemplarily, the target storage method can adopt any one of the blockchain storage strategy, the server storage strategy, the local database storage strategy, etc. Among them, the blockchain storage strategy is preferably used as the above target storage strategy. By using the blockchain storage method to store and manage the target risk assessment model, the immutability and data security of the target risk assessment model can be effectively guaranteed.

[0155] A storage module performs storage processing on the target risk assessment model based on the target storage strategy.

[0156] In this embodiment, the target storage medium corresponding to the target storage policy may be determined, and then the target risk assessment model may be stored in the target storage medium to complete the storage process of the target risk assessment model.

[0157] After screening out the target risk assessment model from all the risk assessment models, the present application will also intelligently use the target storage strategy associated with the target risk assessment model to store the target risk assessment model, which can effectively ensure the data security and non-tamperability of the target risk assessment model, and improve the storage standardization and storage intelligence of the target risk assessment model.

[0158] This application can be applied in the field of smart medical care, thereby promoting the construction of smart cities.

[0159] It should be emphasized that in order to further ensure the privacy and security of the above risk assessment results, the above risk assessment results can also be stored in a blockchain node.

[0160] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0161] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0162] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0163] To solve the above technical problems, the present application also provides a computer device. Figure 3 , Figure 3 This is a basic structural block diagram of the computer device in this embodiment.

[0164] The computer device 3 includes a memory 31, a processor 32, and a network interface 33 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 3 having components 31-33, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0165] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0166] The memory 31 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 31 can be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 31 can also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 3. Of course, the memory 31 can also include both the internal storage unit of the computer device 3 and its external storage device. In this embodiment, the memory 31 is generally used to store the operating system and various application software installed on the computer device 3, such as computer-readable instructions of a risk assessment device based on artificial intelligence. In addition, the memory 31 can also be used to temporarily store various types of data that have been output or are to be output.

[0167] The processor 32 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 32 is generally used to control the overall operation of the computer device 3. In this embodiment, the processor 32 is used to run computer-readable instructions or process data stored in the memory 31, such as computer-readable instructions for running the risk assessment device based on artificial intelligence.

[0168] The network interface 33 may include a wireless network interface or a wired network interface. The network interface 33 is generally used to establish a communication connection between the computer device 3 and other electronic devices.

[0169] In this embodiment, the functions of each module in the risk assessment device based on artificial intelligence as in the above embodiment are realized when the processor executes the computer-readable instructions stored in the memory.

[0170] In this embodiment, the artificial intelligence-based risk assessment device first obtains the target voice data of the target user; then performs audio conversion processing on the target voice data to obtain a corresponding spectrogram; then converts the spectrogram based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrogram; subsequently calls multiple pre-trained risk assessment models; further filters out a target risk assessment model from all the risk assessment models based on preset screening rules; finally, evaluates the embedding vector based on the target risk assessment model to generate a risk assessment result corresponding to the target user. The present application obtains a corresponding embedding vector by performing audio conversion and conversion processing on the acquired target voice data of the target user, and then screens out a target risk assessment model from all pre-trained risk assessment models based on screening rules, and subsequently uses the target risk assessment model to perform evaluation processing on the embedding vector to generate a risk assessment result corresponding to the target user. Different from the existing method in which doctors give risk assessment results for the acquired electroencephalograms based on their own medical experience, the present application performs risk assessment processing on the target voice data of the target user based on the use of a target risk assessment model, which can effectively improve the efficiency of risk assessment for determining whether a user has a risk of suffering from Alzheimer's disease, and improve the accuracy of the generated risk assessment results.

[0171] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor so that the at least one processor performs the functions of each module in the above-mentioned artificial intelligence-based risk assessment device.

[0172] In this embodiment, the artificial intelligence-based risk assessment device first obtains the target voice data of the target user; then performs audio conversion processing on the target voice data to obtain a corresponding spectrogram; then converts the spectrogram based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrogram; subsequently calls multiple pre-trained risk assessment models; further filters out a target risk assessment model from all the risk assessment models based on preset screening rules; finally, evaluates the embedding vector based on the target risk assessment model to generate a risk assessment result corresponding to the target user. The present application obtains a corresponding embedding vector by performing audio conversion and conversion processing on the acquired target voice data of the target user, and then screens out a target risk assessment model from all pre-trained risk assessment models based on screening rules, and subsequently uses the target risk assessment model to perform evaluation processing on the embedding vector to generate a risk assessment result corresponding to the target user. Different from the existing method in which doctors give risk assessment results for the acquired electroencephalograms based on their own medical experience, the present application performs risk assessment processing on the target voice data of the target user based on the use of a target risk assessment model, which can effectively improve the efficiency of risk assessment for determining whether a user has a risk of suffering from Alzheimer's disease, and improve the accuracy of the generated risk assessment results.

[0173] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0174] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A risk assessment device based on artificial intelligence, It is characterized in that include: A first acquisition module, used to acquire target voice data of a target user; A first processing module, used for performing audio conversion processing on the target speech data to obtain a corresponding spectrogram; A second processing module, configured to perform conversion processing on the spectrum graph based on a preset conversion strategy to obtain an embedding vector corresponding to the spectrum graph; A calling module, used to call a plurality of pre-trained risk assessment models; wherein the plurality of risk assessment models are models generated by training a plurality of pre-trained models having different pre-trained weights respectively using a sample data set consisting of pre-collected speech data related to Alzheimer's disease; A screening module, used to screen out a target risk assessment model from all the risk assessment models based on preset screening rules; An evaluation module is used to evaluate the embedding vector based on the target risk evaluation model to generate a risk evaluation result corresponding to the target user.

2. The artificial intelligence-based risk assessment device according to claim 1, It is characterized in that The risk assessment device based on artificial intelligence also includes: A second acquisition module is used to acquire pre-collected speech data related to Alzheimer's disease; an enhancement module, configured to perform an acoustic enhancement operation on the voice data to obtain corresponding first voice data; A labeling module, used for performing data labeling processing on the first voice data to obtain corresponding second voice data; A loading module, used to load a specified pre-trained model corresponding to the specified pre-trained weights; wherein the specified pre-trained model is any one of all the pre-trained models; A conversion module, used for converting the second speech data to obtain a sample data set; A partitioning module, used to divide the sample data set into a training set, a validation set and a test set; The training module is used to train the specified pre-trained model using the training set, and to fine-tune the specified pre-trained model using a preset optimization algorithm to obtain a trained first risk assessment model.

3. The artificial intelligence-based risk assessment device according to claim 2, It is characterized in that The enhancement module comprises: A first processing submodule, configured to perform time distortion processing on the voice data to obtain corresponding third voice data; A second processing submodule is used to perform frequency domain masking processing on the voice data to obtain corresponding fourth voice data; A third processing submodule performs time domain masking processing on the voice data to obtain corresponding fifth voice data; A construction submodule is used to construct the first voice data based on the third voice data, the fourth voice data and the fifth voice data.

4. The artificial intelligence-based risk assessment device according to claim 2, It is characterized in that The screening module comprises: A verification submodule, used to perform performance verification processing on each of the risk assessment models using the verification set to obtain a performance indicator value of each of the risk assessment models corresponding to a preset performance indicator; A first calculation submodule, used to calculate the performance evaluation score of each risk evaluation model based on the performance indicator value of each risk evaluation model; A screening submodule, used to screen out a second risk assessment model with the highest performance assessment score from all the risk assessment models; A first testing submodule, configured to test the second risk assessment model based on the test set to determine whether the second risk assessment model meets preset performance expectations; The first determining submodule is configured to, if yes, use the second risk assessment model as the target risk assessment model.

5. The artificial intelligence-based risk assessment device according to claim 4, It is characterized in that The performance indicators at least include a first performance indicator, a second performance indicator and a third performance indicator; The first calculation submodule includes: A first acquisition unit is used to acquire first performance indicator data corresponding to the first performance indicator, second performance indicator data corresponding to the second performance indicator, and third performance indicator data corresponding to the third performance indicator from the performance indicator data of a specified risk assessment model; wherein the specified risk assessment model is any one of all the risk assessment models; A second acquisition unit, configured to acquire a first weight corresponding to the first performance indicator, a second weight corresponding to the second performance indicator, and a third weight corresponding to the third performance indicator; A calling unit, used to call a preset calculation formula; a calculation unit, configured to calculate and process the first performance indicator data, the second performance indicator data, the third performance indicator data, the first weight, the second weight, and the third weight based on the calculation formula to obtain a corresponding calculation result; A determination unit is used to use the calculation result as a performance evaluation score of the specified risk assessment model.

6. The artificial intelligence-based risk assessment device according to claim 4, It is characterized in that The screening module further includes: An acquisition submodule, configured to acquire a preset model optimization strategy if the second risk assessment model does not meet the performance expectation; An optimization submodule, used to optimize the second risk assessment model based on the model optimization strategy to obtain a corresponding optimization model; A second testing submodule is used to test the optimization model using the test set to obtain a corresponding test performance indicator value; A second calculation submodule, configured to calculate a specified performance evaluation score of the optimization model based on the test performance indicator; A judgment submodule, used to judge whether the specified performance evaluation score is greater than a preset score threshold; The second determining submodule is used to, if yes, use the optimization model as the target risk assessment model.

7. The artificial intelligence-based risk assessment device according to claim 1, It is characterized in that The second processing module comprises: The calling submodule is used to call the pre-trained convolutional neural network; An input submodule, used for inputting the spectrum graph into the convolutional neural network, processing the spectrum graph by the convolutional neural network, and obtaining an output result corresponding to the spectrum graph; The third determination submodule is used to use the output result as the embedding vector.

8. The artificial intelligence-based risk assessment device according to claim 1, It is characterized in that The risk assessment device based on artificial intelligence also includes: A determination module, used to determine a target storage strategy corresponding to the target risk assessment model; A storage module performs storage processing on the target risk assessment model based on the target storage strategy.

9. A computer device, It is characterized in that It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, it implements the functions of each module in the artificial intelligence-based risk assessment device as described in any one of claims 1 to 8.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the functions of each module in the artificial intelligence-based risk assessment device as described in any one of claims 1 to 8.

Images