Swallowing training instrument-based swallowing effect prediction method and system

Abstract

The invention provides a swallowing effect prediction method and system based on a swallowing training instrument, and the method comprises the steps: obtaining swallowing training data transmitted by the swallowing training instrument, and carrying out the preprocessing of the swallowing training data, so as to obtain the preprocessing data; performing Fourier transform and data amplification on the preprocessed data to obtain amplified data; performing high-frequency and low-frequency decomposition denoising processing on the amplified data to obtain denoised data; performing abnormal data identification on the de-noised data by adopting an improved particle swarm optimization algorithm and a density clustering algorithm to obtain target data; the template training data is obtained, the template training data is input into the preset prediction model for training, the target data is input into the trained preset prediction model for effect prediction, and a swallowing effect prediction result is output. According to the swallowing effect prediction method and device, the characteristics of original data can be reserved, low-quality data can be effectively removed, and therefore the final prediction precision of the model is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of effect prediction, and particularly relates to a method and system for predicting swallowing effect based on a swallowing trainer. Background Art

[0002] A swallowing trainer is a medical device specifically used to help patients improve and restore their swallowing function. It generally includes components such as a piezoresistive pressure sensor, a laryngeal muscle trainer, and a controller. The piezoresistive pressure sensor is placed in the patient's throat to detect the movement strength of the patient's laryngeal muscles and generate an electrical signal representing the movement strength. The laryngeal muscle trainer is placed in the throat when the patient is performing swallowing training to conduct targeted training on the patient's laryngeal muscles. The controller is responsible for receiving the electrical signal and controlling the operation of the laryngeal muscle trainer after receiving a swallowing training instruction. In addition, the controller also uploads the electrical signal to the host computer for display, for professional treatment doctors to view and analyze.

[0003] For the recognition of the general swallowing training effect, it is usually judged based on the doctor's experience, which has great limitations. In the prior art, there are also predictions of swallowing effects assisted by a prediction model. However, in the actual prediction process, the data transmitted by the sensor generally includes the patient's current swallowing function data, data of the laryngeal muscles during swallowing training, electromyogram signal data, etc. For the original data, affected by the environment, there will be a large amount of noise in the collected signal data, which will affect the accuracy of swallowing effect prediction. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method and system for predicting swallowing effect based on a swallowing trainer to solve the technical problems in the prior art.

[0005] On the one hand, the present invention provides the following technical solution. A method for predicting swallowing effect based on a swallowing trainer includes: Obtaining swallowing training data transmitted by the swallowing trainer, and preprocessing the swallowing training data to obtain preprocessed data; Performing Fourier transform and data augmentation on the preprocessed data to obtain augmented data; Performing high-frequency and low-frequency decomposition denoising processing on the augmented data to obtain denoised data; Using an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data; Obtaining template training data, inputting the template training data into a preset prediction model for training, and inputting the target data into the trained preset prediction model for effect prediction to output a swallowing effect prediction result.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention first obtains the swallowing training data transmitted by the swallowing training instrument, preprocesses the swallowing training data to obtain preprocessed data; then performs Fourier transform and data augmentation on the preprocessed data to obtain augmented data; then performs high-frequency and low-frequency decomposition denoising processing on the augmented data to obtain denoised data; then uses an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data; finally, obtains template training data, inputs the template training data into a preset prediction model for training, and inputs the target data into the trained preset prediction model for effect prediction to output a swallowing effect prediction result. By performing data augmentation, denoising, and abnormal identification processing on the data, the present invention avoids the problem that too few sample data affects the prediction accuracy of the model, can effectively remove the noise existing in the data while retaining the characteristics of the original data, and finally can effectively eliminate low-quality data, thereby improving the final prediction accuracy of the model.

[0007] Preferably, the step of performing Fourier transform and data augmentation on the preprocessed data to obtain augmented data includes: Performing a fast Fourier transform on the preprocessed data and adding zeros during the fast Fourier transform to obtain transformed data : ; In the formula, is the frequency resolution, is the sampling rate, is the number of sampling points for the fast Fourier transform; Converting the form of the transformed data to obtain a real part and an imaginary part : ; In the formula, is the imaginary symbol; Adding random perturbations to the real part and the imaginary part to obtain a first perturbed part and a second perturbed part : ; ; ; In the formula, , respectively represent the lower and upper limits of the frequency of the transformed data, represents the number of groups of the transformed data; Combining the first perturbed part and the second perturbed part Combine them to obtain combined data, perform a fast inverse Fourier transform on the combined data to obtain restored data, and add the first several data of the restored data to the preprocessed data to obtain intermediate amplified data; Based on the intermediate amplified data, the real part and the imaginary part determine the amplified data.

[0008] Preferably, the step of determining the amplified data based on the intermediate amplified data, the real part and the imaginary part includes: Based on the real part and the imaginary part calculate the data phase : ; Replace the data phases of several groups of random transform data with random values between to obtain an adjusted phase ; Based on the adjusted phase determine the adjusted data : ; In the formula, is the fast inverse Fourier transform; Add the first several data of the adjusted data to the intermediate amplified data to obtain the amplified data.

[0009] Preferably, the step of performing high-frequency and low-frequency decomposition denoising processing on the amplified data to obtain denoised data includes: Use a preset algorithm to decompose the amplified data into several components and a residual component; Set the contribution function of the amplified data: ; In the formula, represents the function variable, is the amplified data, is the solution interval, is the time, is the function value; Determine the distribution characteristics of the contribution function on the coordinate axis and determine the solution interval where the function values are concentrated as the final interval ; Based on the final interval calculate the correlation degree of each component : ; In the formula, represents the th component; Determine the denoised data based on the degree of association.

[0010] Preferably, the step of determining the denoised data based on the degree of association includes: Set a first degree-of-association threshold and a second degree-of-association threshold, and store the components with a degree of association not less than the first degree-of-association threshold into the reconstructed component set, where the first degree-of-association threshold is greater than the second degree-of-association threshold; Store the components with a degree of association not less than the second degree-of-association threshold and less than the first degree-of-association threshold into the denoised component set; Determine the mean square error , and perform denoising smoothing on the components in the denoised component set according to the mean square error to obtain the processed components; ; In the formula, is the polynomial coefficient, is the power of the polynomial, is the number of components in the denoised component set, is the th component in the denoised component set; Combine the processed components, the reconstructed component set, and the residual components for signal reconstruction to obtain the denoised data.

[0011] Preferably, the step of using the improved particle swarm optimization algorithm and the density clustering algorithm to identify abnormal data in the denoised data to obtain the target data includes: Take the neighborhood radius and the minimum number of points included in the neighborhood in the density clustering algorithm as two optimization targets of the improved particle swarm optimization algorithm, and the neighborhood radius and the minimum number of points included in the neighborhood correspond to the velocity and position of the particle respectively; Determine the initial position and initial velocity of each particle in the improved particle swarm optimization algorithm, and perform initial clustering on the denoised data according to the initial position and the initial velocity using the density clustering algorithm; Calculate the fitness of each particle during the clustering process: ; Wherein, represents the average distance between the -th sample in the denoised data and other samples within the same cluster, represents the average distance between the -th sample in the denoised data and samples in other clusters; Iteratively update the inertia weight of each particle : ; Wherein, represents the current iteration number, represents the total iteration number, , respectively represent the starting weight and the final weight; Iteratively update the first learning factor and the second learning factor : ; ; Wherein, respectively represent the starting value and the final value of the first learning factor, respectively represent the starting value and the final value of the second learning factor; Determine the target data according to the fitness, the updated inertia weight , the updated first learning factor and the updated second learning factor .

[0012] Preferably, the step of determining the target data according to the fitness, the updated inertia weight , the updated first learning factor and the updated second learning factor includes: Iteratively update the velocity and position of each particle according to the fitness, the inertia weight , the first learning factor and the second learning factor : ; ; Wherein, respectively represent the velocity of the particle after the -th and -th iterations, respectively represent the position of the particle after the -th and -th iterations, respectively represent the first random number and the second random number, respectively represent the individual optimal solution and the population optimal solution; Repeat the iterative process until the iterative stop condition is met and output the updated neighborhood radius obtained by the final iteration and the minimum number of points included in the updated neighborhood; Based on the updated neighborhood radius and the minimum number of points included in the updated neighborhood, perform density clustering on the denoised data using the density clustering algorithm to obtain the target data.

[0013] In a second aspect, the present invention provides the following technical solution. A swallowing effect prediction system based on a swallowing trainer, the system includes: A processing module, configured to obtain swallowing training data transmitted by the swallowing trainer, and preprocess the swallowing training data to obtain preprocessed data; An amplification module, configured to perform Fourier transform and data amplification on the preprocessed data to obtain amplified data; A denoising module, configured to perform high-frequency and low-frequency decomposition denoising processing on the amplified data to obtain denoised data; An identification module, configured to use an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data; A prediction module, configured to obtain template training data, input the template training data into a preset prediction model for training, and input the target data into the trained preset prediction model for effect prediction to output a swallowing effect prediction result.

[0014] In a third aspect, the present invention provides the following technical solution. A computer includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the swallowing effect prediction method based on a swallowing trainer as described above.

[0015] In a fourth aspect, the present invention provides the following technical solution. A storage medium stores a computer program, and when the computer program is executed by a processor, it implements the swallowing effect prediction method based on a swallowing trainer as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1Flowchart of the swallowing effect prediction method based on a swallowing trainer provided in the first embodiment of the present invention; Figure 2 Block diagram of the structure of the swallowing effect prediction system based on a swallowing trainer provided in the second embodiment of the present invention; Figure 3 Schematic diagram of the hardware structure of a computer provided in another embodiment of the present invention.

[0018] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Specific embodiments

[0019] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the embodiments of the present invention and should not be construed as limiting the present invention.

[0020] Embodiment 1 In the first embodiment of the present invention, as Figure 1 shown, a swallowing effect prediction method based on a swallowing trainer includes: S1. Obtain the swallowing training data transmitted by the swallowing trainer, and preprocess the swallowing training data to obtain preprocessed data; Among them, the swallowing training data here includes current swallowing function data, data of laryngeal muscles during swallowing training, electromyogram signal data, etc. And the preprocessing process here includes format conversion, smoothing processing, etc. And the above preprocessing process is a common step in the prior art. And the swallowing trainer here is a commonly used instrument in the prior art. Therefore, it will not be elaborated here.

[0021] S2. Perform Fourier transform and data augmentation on the preprocessed data to obtain augmented data; Among them, step S2 includes: S21. Perform a fast Fourier transform on the preprocessed data, and add zeros during the fast Fourier transform process to obtain transformed data : ; In the formula, is the frequency resolution, is the sampling rate, is the number of sampling points used for the fast Fourier transform; Specifically, through the fast Fourier transform operation, the discrete time-domain data of each channel can be converted into discrete frequency-domain data.

[0022] S22. Perform a form conversion on the transformed data to obtain the real part and the imaginary part : ; wherein is the imaginary symbol

[0023] S23. Add random perturbations to the real part and the imaginary part to obtain a first perturbed part and a second perturbed part : ; ; ; wherein , respectively represent the lower limit and the upper limit of the frequency of the transformed data, represents the number of groups of the transformed data; Specifically, the purpose of adding the random perturbations here is to mask the noise. By adding several zeros in the process of the fast Fourier transform, , become the original . At the same time, due to the symmetry of the frequency domain data as the node center, the th data of the real part and the imaginary part can be equivalent to the th data after adding the perturbations

[0024] S24. Combine the first perturbed part and the second perturbed part to obtain combined data. Perform an inverse fast Fourier transform on the combined data to obtain restored data, and add the first several data of the restored data to the preprocessed data to obtain intermediate amplified data; S25. Determine amplified data based on the intermediate amplified data, the real part and the imaginary part ; wherein, step S25 includes: S251. Calculate the data phase based on the real part and the imaginary part : .

[0025] S252. Replace the data phases of several random groups of transformed data with random values between to obtain an adjusted phase ; Specifically, for each set of data, there is a data phase. Therefore, randomly select several sets of data and select a random value between them to replace the original phase, and the adjusted phase can be obtained. For those not replaced, their original phases are used.

[0026] S253. Based on the adjusted phase determine the adjusted data : ; wherein, is the fast inverse Fourier transform.

[0027] S254. Add the first several data of the adjusted data to the intermediate amplified data to obtain the amplified data.

[0028] S3. Perform high-frequency and low-frequency decomposition denoising processing on the amplified data to obtain the denoised data; wherein, the step S3 includes: S31. Use a preset algorithm to decompose the amplified data into several component and residual components; Specifically, the preset algorithm here is the adaptive noise complete ensemble empirical mode decomposition algorithm.

[0029] S32. Set the contribution function of the amplified data: ; wherein, represents the function variable, is the amplified data, is the solution interval, is the time, is the function value.

[0030] S33. Determine the distribution characteristics of the contribution function on the coordinate axis and determine the solution interval where the function values are concentrated as the final interval ; Specifically, the contribution function can reflect the contribution degree of the data at different time scales. At the same time, according to the contribution function, due to the random characteristics of the noise, the maximum value of the contribution function is distributed near 0, and the remaining function values except 0 are in a slow decay process and will not decay to a very small value. Therefore, for its function values, they are concentrated near 0. Therefore, in this application, the final interval is set to .

[0031] S34. Based on the final interval calculate the correlation degree of each component : ; In the formula, represents the th component.

[0032] S35. Determine the denoised data based on the correlation degree; Among them, the step S35 includes: S351. Set a first correlation degree threshold and a second correlation degree threshold, and store the components with a correlation degree not less than the first correlation degree threshold into the reconstructed component set, where the first correlation degree threshold is greater than the second correlation degree threshold; Specifically, the first correlation degree threshold here is 0.001, and the second correlation degree threshold is 0.0005. By setting the first correlation degree threshold, the components related to noise and the components related to signal characteristics can be classified, and the purpose of setting the second correlation degree threshold is to eliminate the components with a large amount of noise.

[0033] S352. Store the components with a correlation degree not less than the second correlation threshold and less than the first correlation threshold into the denoised component set.

[0034] S353. Determine the mean square error , and perform denoising and smoothing processing on the components in the denoised component set according to the mean square error to obtain the processed components; ; In the formula, is the polynomial coefficient, is the power of the polynomial, is the number of components in the denoised component set, is the th component in the denoised component set; Specifically, for the denoised component set, the serial number of the middle component is recorded as 0, that is, the above denoising process is based on The sum of the squares of the components centered on is the smallest, that is, the denoising process is completed by the polynomial least squares method.

[0035] S354. Combine and perform signal reconstruction on the processed components, the reconstructed component set, and the residual components to obtain the denoised data.

[0036] S4. Use the improved particle swarm optimization algorithm and density clustering algorithm to identify abnormal data in the denoised data to obtain the target data; Among them, the step S4 includes: S41. Take the neighborhood radius and the minimum number of points included in the neighborhood in the density clustering algorithm as two optimization targets for the improved particle swarm optimization algorithm. The neighborhood radius and the minimum number of points included in the neighborhood correspond to the velocity and position of the particle respectively; Specifically, the density clustering algorithm here is a commonly used clustering algorithm in the prior art. Replace the two most important parameters in the density clustering algorithm with the velocity and position of the particle in the improved particle swarm optimization algorithm, that is, in the process of density clustering, use the improved particle swarm optimization algorithm to perform iterative update and optimization of the parameters.

[0037] S42. Determine the initial position and initial velocity of each particle in the improved particle swarm optimization algorithm, and perform initial clustering on the denoised data according to the initial position and the initial velocity by using the density clustering algorithm.

[0038] S43. Calculate the fitness of each particle during the clustering process : ; In the formula, represents the average distance between the th sample in the denoised data and other samples in the same cluster, represents the average distance between the th sample in the denoised data and samples in other clusters; Specifically, the purpose of calculating the fitness is to ensure that the position of the particle optimizes towards the maximum value of the fitness.

[0039] S44. Iteratively update the inertia weight of each particle : ; In the formula, represents the current iteration number, represents the total iteration number, , respectively represent the starting weight and the final weight; Specifically, the purpose of updating the weight is to have a greater global optimization ability in the early stage of the algorithm operation and a greater local convergence ability in the later stage.

[0040] S45. Iteratively update the first learning factor and the second learning factor : ; ; In the formula, respectively represent the starting value and the final value of the first learning factor, respectively represent the starting value and the final value of the second learning factor; Specifically, the first learning factor and the second learning factor here are asymmetric learning factors. In the early stage, a larger first learning factor and a smaller second learning factor are adopted to make the particles pay more attention to the individual optimal solution and reduce the influence of the cluster. This can make the particles cover the entire search space as much as possible. As the number of iterations increases, the first learning factor decreases linearly and the second learning factor increases linearly, thereby strengthening the convergence ability of the particles to the global optimal point.

[0041] S46. Determine the target data according to the fitness, the updated inertia weight the updated first learning factor and the updated second learning factor ; Among them, the step S46 includes: S461. Iteratively update the velocity and position of each particle according to the fitness, the inertia weight the first learning factor and the second learning factor : ; ; In the formula, respectively represent the velocity of the particle after the -th and -th iterations, respectively represent the position of the particle after the -th and -th iterations, respectively represent the first random number and the second random number, respectively represent the individual optimal solution and the global optimal solution; Among them, the first random number and the second random number here are specifically random numbers between 0 and 1.

[0042] S462. Repeat the iterative process until the iteration stop condition is satisfied, and output the updated neighborhood radius and the minimum number of points included in the updated neighborhood obtained by the final iteration.

[0043] S463. Perform density clustering on the denoised data based on the updated neighborhood radius and the minimum number of points included in the updated neighborhood by using the density clustering algorithm to obtain the target data; Specifically, in the process of density clustering, several clustering clusters will be obtained, and there will be some outliers outside the clustering clusters, and these outliers are abnormal data. Only the data in the clustering clusters are retained to obtain the target data.

[0044] S5, obtaining template training data, inputting the template training data into a preset prediction model for training, inputting the target data into the trained preset prediction model for effect prediction, and outputting a swallowing effect prediction result; Specifically, the preset prediction model here is specifically a BP neural network model in the prior art. By inputting the target data into the trained model, the final swallowing effect prediction result can be output.

[0045] The swallowing effect prediction method based on the swallowing training instrument provided in the first embodiment of the present invention first obtains the swallowing training data transmitted by the swallowing training instrument, and pre-processes the swallowing training data to obtain pre-processed data; then performs Fourier transform and data amplification on the pre-processed data to obtain amplified data; then performs high-frequency and low-frequency decomposition and denoising processing on the amplified data to obtain denoised data; then uses the improved particle swarm optimization algorithm and the density clustering algorithm to identify abnormal data on the denoised data to obtain target data; finally, obtains template training data, inputs the template training data into a preset prediction model for training, and inputs the target data into the trained preset prediction model for effect prediction, so as to output the swallowing effect prediction result. The present invention avoids the problem of too little sample data affecting the prediction accuracy of the model by amplifying, denoising and identifying abnormalities in the data, can effectively remove the noise in the data while retaining the characteristics of the original data, and finally can effectively eliminate low-quality data, thereby improving the final prediction accuracy of the model.

[0046] Embodiment 2 like Figure 2 As shown, in the second embodiment of the present invention, a swallowing effect prediction system based on a swallowing training device is provided, and the system includes: Processing module 1, used for acquiring swallowing training data transmitted by the swallowing training device, and preprocessing the swallowing training data to obtain preprocessed data; Amplification module 2, used for performing Fourier transform and data amplification on the pre-processed data to obtain amplified data; A denoising module 3 is used to perform high-frequency and low-frequency decomposition and denoising processing on the amplified data to obtain denoised data; Identification module 4, used to identify abnormal data on the denoised data using an improved particle swarm optimization algorithm and a density clustering algorithm to obtain target data; Prediction module 5, used for acquiring template training data, inputting the template training data into a preset prediction model for training, inputting the target data into the trained preset prediction model for effect prediction, and outputting a swallowing effect prediction result; The amplification module 2 comprises: The transformation sub-module is used to perform a fast Fourier transform on the preprocessed data and add zeros during the fast Fourier transform to obtain transformed data : ; In the formula, is the frequency resolution, is the sampling rate, is the number of sampling points for the fast Fourier transform; The conversion sub-module is used to perform a form conversion on the transformed data to obtain the real part and the imaginary part : ; In the formula, is the imaginary symbol; The addition sub-module is used to add random perturbations to the real part and the imaginary part to obtain the first perturbed part and the second perturbed part : ; ; ; In the formula, , respectively represent the lower limit and upper limit of the frequency of the transformed data, represents the number of groups of the transformed data; The intermediate amplification sub-module is used to combine the first perturbed part and the second perturbed part to obtain combined data, perform an inverse fast Fourier transform on the combined data to obtain restored data, and add the first several data of the restored data to the preprocessed data to obtain intermediate amplified data; The final amplification sub-module is used to determine the amplified data based on the intermediate amplified data, the real part and the imaginary part .

[0047] The final amplification sub-module includes: The phase unit is used to calculate the data phase based on the real part and the imaginary part : ; The replacement unit is used to replace the data phases of several random groups of transformed data with random values between to obtain the adjusted phase ; An adjustment unit, configured to adjust the phase based on the Determine adjustment data : ; In the formula, is the fast inverse Fourier transform; The amplification unit is used to add the first several data of the adjustment data to the intermediate amplified data to obtain the amplified data.

[0048] The denoising module 3 comprises: The decomposition submodule is used to decompose the amplification data into several Components and residual components; The contribution submodule is used to set the contribution function of the amplification data: ; In the formula, Represents a function variable, To expand the data, To solve the interval, For time, is the function value; The interval submodule is used to determine the distribution characteristics of the contribution function on the coordinate axis and determine the solution interval in the function value set as the final interval ; An association submodule is used to associate the final interval Calculate each of the Correlation of components : ; In the formula, Indicates indivual Quantity; The denoising submodule is used to determine denoised data based on the correlation degree.

[0049] The denoising submodule comprises: The dividing unit is used to set a first correlation threshold and a second correlation threshold, and divide the The components are stored in a reconstructed component set, wherein the first relevance threshold is greater than the second relevance threshold; A denoising component unit is used to convert the correlation degree not less than the second correlation threshold and less than the first correlation threshold The components are stored in the denoising component set; Smoothing unit, used to determine the mean square error , and the denoising component set is sorted according to the mean square error Denoise and smooth the component to obtain the processed component; ; In the formula, is the polynomial coefficient, is the power of the polynomial, is the number of components in the denoised component set, is the th component in the denoised component set; Reconstruction unit, used to combine the processed component, the reconstructed component set and the residual component, and perform signal reconstruction to obtain denoised data.

[0050] The recognition module 4 includes: Sub-module to be optimized, used to use the neighborhood radius and the minimum number of points included in the neighborhood in the density clustering algorithm as two optimization targets of the improved particle swarm optimization algorithm. The neighborhood radius and the minimum number of points included in the neighborhood correspond to the velocity and position of the particle respectively; Initial sub-module, used to determine the initial position and initial velocity of each particle in the improved particle swarm optimization algorithm, and perform initial clustering on the denoised data according to the initial position and the initial velocity using the density clustering algorithm; Fitness sub-module, used to calculate the fitness of each particle during the clustering process : ; In the formula, represents the average distance between the th sample in the denoised data and other samples in the same cluster, represents the average distance between the th sample in the denoised data and samples in other clusters; First update sub-module, used to iteratively update the inertia weight of each particle: ; In the formula, represents the current iteration number, represents the total iteration number, , respectively represent the starting weight and the final weight; Second update sub-module, used to iteratively update the first learning factor and the second learning factor : ; ; In the formula, respectively represent the starting value and the final value of the first learning factor, respectively represent the starting value and the final value of the second learning factor; An output sub-module, configured to determine target data according to the fitness, the updated inertia weight , the updated first learning factor and the updated second learning factor ;

[0051] The output sub-module includes: An update unit, configured to iteratively update the velocity and position of each particle according to the fitness, the inertia weight , the first learning factor and the second learning factor : ; ; In the formula, respectively represent , the velocity of the particle after the respectively represent , position of the particle after the respectively represent the first random number and the second random number, respectively represent the individual optimal solution and the global optimal solution; A parameter output unit, configured to repeat the iterative process until the iteration stop condition is satisfied and output the updated neighborhood radius and the minimum number of points included in the updated neighborhood obtained by the final iteration; A clustering unit, configured to perform density clustering on the denoised data based on the updated neighborhood radius and the minimum number of points included in the updated neighborhood by using a density clustering algorithm to obtain target data.

[0052] In some other embodiments of the present invention, the present invention provides the following technical solution. A computer includes a memory 102, a processor 101, and a computer program stored on the memory 102 and executable on the processor 101. When the processor 101 executes the computer program, the swallowing effect prediction method based on a swallowing trainer as described above is implemented.

[0053] Specifically, the above-mentioned processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits for implementing the embodiments of the present invention.

[0054] Among them, the memory 102 may include a mass storage for data or instructions. By way of example and not limitation, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid state drive (SSD), a flash memory, an optical disc, a magneto-optical disc, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. In suitable cases, the memory 102 may include removable or non-removable (or fixed) media. In suitable cases, the memory 102 may be internal or external to the data processing device. In a particular embodiment, the memory 102 is a non-volatile memory. In a particular embodiment, the memory 102 includes a read-only memory (ROM) and a random access memory (RAM). In suitable cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM), or a flash memory, or a combination of two or more of these. In suitable cases, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM may be a fast page mode dynamic random access memory (FPMDRAM), an extended data out dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0055] The memory 102 can be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 101.

[0056] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-mentioned swallowing effect prediction method based on the swallowing trainer.

[0057] In some of these embodiments, the computer may further include a communication interface 103 and a bus 100. Among them, as Figure 3 shown, the processor 101, the memory 102, and the communication interface 103 are connected through the bus 100 and complete communication with each other.

[0058] The communication interface 103 is used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present invention. The communication interface 103 can also implement data communication with other components, such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations, etc.

[0059] Bus 100 includes hardware, software, or both, and couples components of a computer device to each other. Bus 100 includes, but is not limited to, at least one of the following: Data Bus, Address Bus, Control Bus, Expansion Bus, Local Bus. By way of example and not limitation, Bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable bus or a combination of two or more of these. In suitable cases, Bus 100 may include one or more buses. Although embodiments of the present invention describe and illustrate specific buses, the present invention contemplates any suitable bus or interconnect.

[0060] The computer may execute the swallowing effect prediction method based on a swallowing trainer of the present invention based on the obtained swallowing effect prediction system based on a swallowing trainer, so as to implement the swallowing effect prediction based on a swallowing trainer.

[0061] In still some other embodiments of the present invention, in combination with the above-mentioned swallowing effect prediction method based on a swallowing trainer, embodiments of the present invention provide the following technical solution: a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned swallowing effect prediction method based on a swallowing trainer is implemented.

[0062] Those skilled in the art can understand that the logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus or device), or used in combination with these instruction execution systems, apparatus or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device.

[0063] More specific examples (non-exhaustive list) of readable media include the following: an electrical connection part (electronic device) having one or more wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation or other suitable processing as necessary, and then stored in a computer memory.

[0064] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0065] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity in description, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0066] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. A method for predicting swallowing effect based on a swallowing trainer, characterized in that, Including: Obtain the swallowing training data transmitted by the swallowing training instrument, and preprocess the swallowing training data to obtain preprocessed data; Perform Fourier transform and data augmentation on the preprocessed data to obtain augmented data; Perform high-frequency and low-frequency decomposition denoising processing on the augmented data to obtain denoised data; Use an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data; Obtain template training data, input the template training data into a preset prediction model for training, and input the target data into the trained preset prediction model for effect prediction to output a swallowing effect prediction result.

2. The swallowing effect prediction method based on a swallowing trainer according to claim 1, wherein, The step of performing Fourier transform and data augmentation on the preprocessed data to obtain augmented data includes: Perform a fast Fourier transform on the preprocessed data and add zeros during the fast Fourier transform process to obtain transformed data : ； In the formula, is the frequency resolution, is the sampling rate, is the number of sampling points for the fast Fourier transform; Perform a format conversion on the transformation data to obtain the real part and the imaginary part : ； In the formula, is the imaginary symbol; Add random perturbations to the real part and the imaginary part to obtain a first perturbed part and a second perturbed part : ； ； ； In the formula, , respectively represent the lower limit and the upper limit of the frequency of the transformed data, represents the number of groups of the transformed data; Combine the first perturbation part with the second perturbation part to obtain combined data, perform a fast inverse Fourier transform on the combined data to obtain restored data, and add the first several data of the restored data to the preprocessed data to obtain intermediate amplified data; Based on the intermediate amplification data and the real part and the imaginary part determine the amplification data.

3. The swallowing effect prediction method based on a swallowing trainer according to claim 2, characterized in that Based on the intermediate amplification data and the real part and the imaginary part The steps for determining the amplification data include: Based on the real part and the imaginary part calculate the data phase : ； Replace the data phase of several random sets of transformation data with a random value between to obtain an adjusted phase ; Based on the adjusted phase Determine the adjustment data : ； In the formula, is the fast inverse Fourier transform; Add the first several data of the adjusted data to the intermediate augmented data to obtain augmented data.

4. The swallowing effect prediction method based on a swallowing trainer according to claim 1, wherein, The step of performing high-frequency and low-frequency decomposition denoising processing on the augmented data to obtain denoised data includes: Decompose the amplified data into a number of component and residual components by using a preset algorithm; Set the contribution function of the augmented data: ； In the formula, represents the function variable, is the amplified data, is the solution interval, is the time, is the function value; Determine the distribution characteristics of the contribution function on the coordinate axes and determine the solution interval where the function values are concentrated as the final interval ; Based on the final interval Calculate each of the correlation degrees of the components : ； In the formula, represents the th component; Determine the denoised data based on the correlation degree.

5. The swallowing effect prediction method based on a swallowing trainer according to claim 4, characterized in that, The step of determining the denoised data based on the correlation degree includes: Set a first correlation threshold and a second correlation threshold, and store the components with a correlation not less than the first correlation threshold into the reconstructed component set, where the first correlation threshold is greater than the second correlation threshold; The components are stored in the reconstructed component set, where the first correlation threshold is greater than the second correlation threshold; Store the components with a correlation degree not less than the second correlation threshold and less than the first correlation threshold into the denoised component set; ​ Determine the mean square error , and perform denoising and smoothing processing on the components in the denoised component set according to the mean square error to obtain a processed component; component ； Wherein, is the polynomial coefficient, is the power of the polynomial, is the number of components in the denoised component set, is the th component in the denoised component set; Combine the processed component, the set of reconstructed components, and the residual component, and perform signal reconstruction to obtain denoised data.

6. The swallowing effect prediction method based on a swallowing trainer according to claim 1, wherein The step of using an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data includes: Take the neighborhood radius in the density clustering algorithm and the minimum number of points included in the neighborhood as two optimization objectives for improving the particle swarm optimization algorithm. The neighborhood radius and the minimum number of points included in the neighborhood correspond to the velocity and position of the particle respectively; Determine the initial position and initial velocity of each particle in the improved particle swarm optimization algorithm, and perform initial clustering on the denoised data according to the initial position and the initial velocity by using a density clustering algorithm; Calculate the fitness of each particle during the clustering process : ； In the formula, represents the average distance between the -th sample in the denoised data and other samples within the same cluster, represents the average distance between the -th sample in the denoised data and samples in other clusters; Iteratively update the inertia weight of each particle : ； In the formula, represents the current iteration number, represents the total number of iterations, , represent the starting weight and the final weight respectively; Iteratively update the first learning factor and the second learning factor : ； ； wherein, respectively represent the starting value and the final value of the first learning factor, respectively represent the starting value and the final value of the second learning factor; Determine the target data according to the fitness and the updated inertia weight , the updated first learning factor and the updated second learning factor .

7. The swallowing effect prediction method based on a swallowing trainer according to claim 6, characterized in that, The steps of determining the target data according to the fitness and the updated inertia weight , the updated first learning factor and the updated second learning factor include: According to the fitness and inertia weight , the first learning factor and the second learning factor iterate and update the velocity and position of each particle: ； ； In the formula, respectively represent the velocity of the particle after the -th and -th iterations, respectively represent the position of the particle after the -th and -th iterations, respectively represent the first random number and the second random number, respectively represent the individual optimal solution and the global optimal solution; Repeat the iteration process until the iteration stop condition is met, and output the updated neighborhood radius and the minimum number of points included in the updated neighborhood obtained by the final iteration; Perform density clustering on the denoised data based on the updated neighborhood radius, the minimum number of points included in the updated neighborhood, and by using a density clustering algorithm to obtain target data.

8. A swallowing effect prediction system based on a swallowing trainer, characterized in that, The system includes: A processing module, configured to obtain the swallowing training data transmitted by the swallowing training instrument, and preprocess the swallowing training data to obtain preprocessed data; An augmentation module, configured to perform Fourier transform and data augmentation on the preprocessed data to obtain augmented data; A denoising module, configured to perform high-frequency and low-frequency decomposition denoising processing on the augmented data to obtain denoised data; An identification module, configured to use an improved particle swarm optimization algorithm and a density clustering algorithm to identify abnormal data in the denoised data to obtain target data; A prediction module, configured to obtain template training data, input the template training data into a preset prediction model for training, and input the target data into the trained preset prediction model for effect prediction to output a swallowing effect prediction result.

9. A computer, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the swallowing effect prediction method based on a swallowing training instrument according to any one of claims 1 to 7.

10. A storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the swallowing effect prediction method based on a swallowing training instrument according to any one of claims 1 to 7.

Images