Sole health detection system and method based on flexible capacitive sensor
By integrating flexible capacitive sensors and neural network models on smart insoles, processing and predicting plantar pressure data, the problems of plantar pressure detection accuracy and sensor volume limitation in the existing technology are solved, and efficient and accurate plantar health detection is achieved .
Patent Information
- Application Number
- CN202411289579.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-11-08
AI Technical Summary
The existing plantar pressure testing methods, such as piezoresistive sensors and plantar pressure testing plates, have problems such as temperature changes affecting measurement accuracy and excessive sensor volume leads to insufficient number of sensors per unit area in the insole, which in turn affects the pressure acquisition effect.
The smart insole based on flexible capacitive sensors is used to collect sole pressure data, and the data processing and prediction are carried out through the SVR-LSTM neural network model and the improved Boxer Squid algorithm optimized BP neural network, combining data preprocessing, feature extraction and normalization Processing, establish a two-dimensional vector data set to improve the accuracy of plantar pressure detection.
It achieves higher accuracy and reliability of plantar pressure detection, can more comprehensively process the nonlinear and time series characteristics of plantar pressure data, provide a more comfortable detection experience, and improve the efficiency and accuracy of the detection system.
Smart Images

Figure CN118902434A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of foot health detection technology, and in particular to a foot health detection system and method based on a flexible capacitive sensor. Background Art
[0002] With the development of society, people's requirements for quality of life are constantly increasing, and a healthy body is undoubtedly the beginning of a beautiful and happy life. Standardized and parameterized measurement of various physiological indicators of the body and the development of good living habits have gradually become a consensus. Studies have shown that the distribution information of plantar pressure during walking is an important basis for the diagnosis of abnormal foot morphology. High plantar pressure is an important risk factor for foot development. Therefore, accurate measurement of plantar pressure is very important.
[0003] The commonly used plantar pressure testing methods now include piezoresistive sensors and plantar pressure testing boards. Piezoresistive sensors are sensitive to temperature changes, which will affect the measurement accuracy of the sensor. The plantar pressure testing board has a large sensor volume, which limits the number of sensors per unit area in the insole, and the pressure collection effect is not ideal.
[0004] In summary, developing a foot health detection system and method based on flexible capacitive sensors is still a key issue that needs to be urgently solved in the field of foot health detection. Summary of the invention
[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a foot health detection system and method based on a flexible capacitive sensor, which has better recognition effect and higher accuracy.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] A method for detecting foot health based on a flexible capacitive sensor comprises the following steps:
[0008] S1. Collect the plantar pressure data of the subject in a free state through a smart insole with a flexible capacitive sensor, and transmit the collected plantar pressure data to a computer via WIFI;
[0009] S2, the computer obtains the original plantar pressure data and performs data preprocessing, feature extraction and normalization, and establishes a two-dimensional vector data set with the processed data value and the real data value;
[0010] S3, divide the data set into training set and test set in proportion;
[0011] S4, constructing an SVR-LSTM neural network model, and inputting the data in the training set into the SVR-LSTM neural network model for the first training;
[0012] S5, inputting the model parameters obtained from the first training into the BP neural network optimized based on the improved Salp Integrity Algorithm to perform the second training;
[0013] S6. Denormalize the plantar pressure prediction result obtained by the second training to obtain a model threshold.
[0014] The present invention is further configured as follows: in step S1, the smart insole with a flexible capacitive sensor has nine collection areas. When the plantar pressure of the subject is collected, nine plantar pressure characteristic areas can be collected, including the first metatarsal, the first metatarsal, the second metatarsal, the third metatarsal, the fourth metatarsal, the fifth metatarsal, the midfoot, the inner side of the heel, the outer side of the heel and the back side of the heel.
[0015] The present invention is further configured as follows: in step S2, the following steps are included:
[0016] S21, preprocessing the acquired plantar pressure data, including using a filter to remove various noises existing in the plantar pressure data;
[0017] S22, when extracting features from the acquired plantar pressure data, including time domain eigenvalues, frequency domain eigenvalues, and nonlinear dynamics eigenvalues;
[0018] S23, normalize the processed data, and the calculation formula is as follows:
[0019]
[0020] In the formula, X' is the normalized value of the original data, X min and X max They are the minimum and maximum values in the original data respectively. After normalization, the original data are mapped to the interval [0, 1].
[0021] The present invention is further configured as follows: in step S3, the data set is a two-dimensional vector (X i ,Y i ), where X i is the processed data, Y i is the original data, that is, the true value of the data.
[0022] The present invention is further configured as follows: in step S4, the following steps are included:
[0023] S41. Build the SVR model and LSTM model, and the SVR model function is as follows:
[0024]
[0025] In the formula, K(x i ,x j ) is the kernel function, a i is the ith Lagrangian coefficient above the upper edge of the isolation zone, is the i-th Lagrangian coefficient below the upper edge of the isolation zone;
[0026] S42, during the first training, the two-dimensional vector (X i ,Y i ) as training samples, and use SVR model and LSTM model for training respectively. The obtained threshold is recorded as SVR(X i ) and LSTM(X i );
[0027] S43, then the two-dimensional vector (SVR (X i ), LSTM(X i )) and the corresponding true value Y i The parameters are input as model parameters into the BP neural network optimized based on the improved salp algorithm, and the BP neural network optimized based on the improved salp algorithm is trained.
[0028] The present invention is further configured as follows: in step S41, the LSTM model includes a memory unit, an input gate, an output gate and a forget gate, the input gate and the output gate are used to receive, output and correct parameters, the forget gate is used to selectively clear specific correction parameters, and the memory unit is used to record and save the current state of the neuron.
[0029] The present invention is further configured as follows: in step S5, the BP neural network optimized based on the improved salp algorithm is to use the weights and thresholds of the BP neural network optimized based on the improved salp algorithm.
[0030] The present invention is further configured as follows: in step S6, the calculation formula of the denormalization process is as follows:
[0031] X″=X′(X max -X min )+X min ;
[0032] In the formula, X' is the normalized data, and X" is the result after denormalization.
[0033] The present invention also provides a foot health detection system based on a flexible capacitive sensor, comprising:
[0034] A data acquisition module, wherein the data acquisition module collects the plantar pressure data of the subject in a free state through a smart insole with a flexible capacitive sensor, and transmits the collected plantar pressure data to the computer module through a data transmission module;
[0035] A data transmission module, the data transmission module is used to transmit the plantar pressure data collected by the data collection module to the computer module;
[0036] A computer module is used to process, divide, train and denormalize the original plantar pressure data to obtain a model threshold.
[0037] The present invention is further configured as follows: the computer module includes a processing module, a partitioning module, a first training module, a second training module, and a denormalization processing module:
[0038] The processing module is used to perform data preprocessing, feature extraction and normalization processing on the original plantar pressure data, and establish a two-dimensional vector data set;
[0039] The partitioning module is used to divide the data set into a training set and a test set in proportion;
[0040] The first training module trains the data in the training set through the SVR-LSTM neural network model to obtain model parameters;
[0041] The second training module trains the model parameters by using a BP neural network optimized based on an improved Salp Intense algorithm to obtain a plantar pressure prediction result;
[0042] The denormalization processing module performs denormalization processing on the plantar pressure prediction result to obtain a model threshold.
[0043] Beneficial Effects
[0044] Compared with the known public technology, the technical solution provided by the present invention has the following advantages:
[0045] Beneficial effects:
[0046] (1) In the present invention, the foot detection system based on the flexible capacitive sensor is composed of a data acquisition module, a data transmission module and a computer module. The data acquisition module collects the foot pressure data of the subject in a free state through an intelligent insole with a flexible capacitive sensor, and transmits the collected foot pressure data to the computer module through the data transmission module. The computer module processes, divides, trains and denormalizes the original foot pressure data to obtain a model threshold. The detection system of the present invention can process the foot pressure data more comprehensively, including its nonlinear characteristics, time series characteristics and prediction tasks that require fine adjustment, thereby improving the accuracy and reliability of foot pressure detection and providing an efficient and accurate solution for foot health detection.
[0047] (2) In the present invention, by using a smart insole with a flexible capacitive sensor, it can better adapt to the shape of the subject's foot and provide a more comfortable detection experience, thereby ensuring that the plantar pressure data collected during the subject's walking process can be closer to the actual situation, making the acquired data more accurate and objective.
[0048] (3) In the present invention, the SVR model can process the complex patterns of plantar pressure data during training, providing more accurate input for the subsequent LSTM model and the BP neural network training based on the improved salp algorithm optimization. The LSTM model processes the dynamic characteristics of plantar pressure changing over time, providing time series insights for the prediction of plantar pressure. The BP neural network optimized based on the improved salp algorithm optimizes the network parameters by simulating the foraging behavior of salps, thereby improving the learning efficiency and prediction accuracy of the network. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 A schematic diagram of the collection area of a smart insole for a foot health detection method based on a flexible capacitive sensor;
[0050] Figure 2 The schematic diagram of a method for detecting foot health based on a flexible capacitive sensor is shown below;
[0051] Figure 3 This is a schematic diagram of an SVR-LSTM neural network model for a foot health detection method based on a flexible capacitive sensor;
[0052] Figure 4 This is a system diagram of a foot health detection system based on flexible capacitive sensors.
[0053] Explanation of the numbers in the figure: 1. Data acquisition module; 2. Data transmission module; 3. Computer module; 31. Processing module; 32. Division module; 33. First training module; 34. Second training module; 35. Denormalization processing module. DETAILED DESCRIPTION
[0054] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0055] The present invention will be further described below in conjunction with the embodiments.
[0056] Example 1
[0057] Please refer to Figure 1 - Figure 3 As shown, a foot health detection method based on a flexible capacitive sensor includes the following steps:
[0058] Step 1: Collect the plantar pressure data of the subject in a free state through a smart insole with a flexible capacitive sensor, and transmit the collected plantar pressure data to a computer via WIFI;
[0059] like Figure 1 As shown, the smart insole with a flexible capacitive sensor has nine collection areas. When the plantar pressure of the subject is collected, nine plantar pressure characteristic areas can be collected, including the first metatarsal, the first metatarsal, the second metatarsal, the third metatarsal, the fourth metatarsal, the fifth metatarsal, the midfoot, the inner side of the heel, the outer side of the heel and the back side of the heel.
[0060] Step 2: The computer obtains the original plantar pressure data and performs data preprocessing, feature extraction and normalization, and establishes a two-dimensional vector data set with the processed data value and the real data value, including the following steps:
[0061] S21, preprocessing the acquired plantar pressure data, including using a filter to remove various noises existing in the plantar pressure data;
[0062] S22, when extracting features from the acquired plantar pressure data, including time domain eigenvalues, frequency domain eigenvalues, and nonlinear dynamics eigenvalues;
[0063] Among them, when extracting features from the acquired plantar pressure data, three time domain eigenvalues (peak value, root mean square value, mean of the absolute value of the first-order difference) and one frequency domain eigenvalue (Welch method) were extracted from each channel, and five nonlinear dynamic eigenvalues (sample entropy, sorting entropy, ENTR in the recursive graph, Hurst index, and Hilbert-Huang spectrum entropy) were combined for feature combination;
[0064] S23, normalize the processed data, and the calculation formula is as follows:
[0065]
[0066] In the formula, X' is the normalized value of the original data, X min and X max They are the minimum and maximum values in the original data respectively. After normalization, the original data are mapped to the interval [0, 1].
[0067] Step 3: Divide the data set into training set and test set in proportion;
[0068] The data set is a two-dimensional vector (X i,Y i ), where X i is the processed data, Y i is the original data, that is, the true value of the data.
[0069] Step 4: Build an SVR-LSTM neural network model and input the data in the training set into the SVR-LSTM neural network model for the first training. Figure 3 As shown, the following steps are included:
[0070] S41. Build the SVR model and LSTM model, and the SVR model function is as follows:
[0071]
[0072] In the formula, K(x i ,x j ) is the kernel function, a i is the ith Lagrangian coefficient above the upper edge of the isolation zone, is the i-th Lagrangian coefficient below the upper edge of the isolation zone;
[0073] Among them, the LSTM model includes a memory unit, an input gate, an output gate, and a forget gate. The input gate and the output gate are used to receive, output, and correct parameters. The forget gate is used to selectively clear specific correction parameters. The memory unit is used to record and save the current state of the neuron.
[0074] The input gate is responsible for deciding how much input information to keep at the current moment The unit state c at the current moment t , consists of two parts. The first part is the sigmoid layer, which determines the value i to be updated t The second part is the tanh layer, which creates a new cell state value vector Update the information that needs to be updated to the newly created cell state value vector inside, Finally, it will be added to the current unit state c t in,i t and The calculation formula is as follows:
[0075] i t =σ(w i ·[h t-1 ,x t ]+b i )
[0076]
[0077] In the formula, w i and w care the weight parameters corresponding to sigmoid and tanh in the input gate, b i and b c is the corresponding bias term, the current cell state c t The calculation formula is as follows:
[0078]
[0079] The output gate is responsible for determining the cell state c at the current moment t How many outputs go to the hidden state h of the unit t , through the sigmoid layer to determine the unit state c t The decision vector o t , the current cell state c t Through tanh processing, a value between (-1, 1) is obtained, and this value and the decision vector o are combined. t Multiply them together to get the final output hidden state h t , the calculation formula is as follows:
[0080] o t =σ(w o ·[h t-1 ,x t ]+b o )
[0081] h t =o t *tanh(c t );
[0082] The forget gate is responsible for deciding how much of the previous cell state c to retain. t-1 The unit state c at the current moment t , that is, decide what information to discard from the cell state. This gate reads the hidden state h from the previous moment. t-1 and the current input x t Connect to form a new feature vector, and after the sigmoid activation function, output a vector f t , f t The value range of each element in is (0, 1), through c t-1 ×f t To determine the unit state c at the previous moment t-1 How many are added to the current cell state c t The result is "0" for complete discarding and "1" for complete retention. t The calculation formula is as follows:
[0083] f t =qσ(w f ·[h t-1 ,xt ]+b f )
[0084] In the formula, w f and b f They are the weight parameters and bias terms corresponding to the sigmoid activation function in the forget gate respectively;
[0085] S42, during the first training, the two-dimensional vector (X i ,Y i ) as training samples, and use SVR model and LSTM model for training respectively. The obtained threshold is recorded as SVR(X i ) and LSTM(X i );
[0086] S43, then the two-dimensional vector (SVR (X i ), LSTM(X i )) and the corresponding true value Y i The parameters are input as model parameters into the BP neural network optimized based on the improved salp algorithm, and the BP neural network optimized based on the improved salp algorithm is trained.
[0087] Step 5: input the model parameters obtained from the first training into the BP neural network optimized based on the improved Salp Intestine algorithm for the second training;
[0088] The BP neural network optimized based on the improved salp algorithm is to use the weights and thresholds of the BP neural network optimized based on the improved salp algorithm;
[0089] Among them, based on the improved Salp Algorithm, the leader searches for the optimal position forward, and the leader's position update formula is as follows:
[0090]
[0091] Generate a scale-free network to map the relationship of followers, and then add the inertia weight based on population diversity feedback to the update formula of the follower salps. The specific formula is as follows:
[0092]
[0093] Where nbd(i) represents all neighbors with which the ith salp has a topological relationship. The smaller the value of the inertia weight W, the more likely the algorithm is to deviate from the local optimum and converge to the optimum. The calculation formula of the inertia weight is as follows:
[0094]
[0095] In the formula, C represents the diversity metric, l maxRepresents the maximum number of iterations, and the diversity metric can be calculated as follows:
[0096]
[0097] In the formula, N refers to the number of individuals in the population, Dim represents the dimension of the variable, represents the j-th dimension variable of the ith individual, Represents the average value of all individuals in the population in the jth dimension, the diagonal distance of the search space:
[0098]
[0099] Among them, in the BP neural network optimized based on the improved Salp Algorithm, a threshold of the BP neural network is set to define the t of the BP neural network. k is the expected output value, and the calculation formula is as follows:
[0100]
[0101] Define the BP neural network output value as z k , the calculation formula is as follows:
[0102]
[0103] If the output is different from the expected output, there is an error e, which is calculated as follows:
[0104]
[0105] If the output error e is greater than E, continue training. If the output error e is less than or equal to E, end network training and input the test sample into the trained network to get the prediction result.
[0106] Step 6: Denormalize the plantar pressure prediction result obtained by the second training to obtain a model threshold;
[0107] Among them, the calculation formula for the denormalization process is as follows:
[0108] X″=X′(X max -X min )+X min ;
[0109] In the formula, X' is the normalized data, and X" is the result after denormalization.
[0110] In this embodiment, in the present invention, the SVR model can process the complex patterns of plantar pressure data during training, providing more accurate input for subsequent LSTM model and BP neural network training based on the improved salp algorithm optimization. The LSTM model processes the dynamic characteristics of plantar pressure changing over time, providing time series insights for the prediction of plantar pressure. The BP neural network optimized based on the improved salp algorithm optimizes network parameters by simulating the foraging behavior of salps, thereby improving the learning efficiency and prediction accuracy of the network.
[0111] Example 2
[0112] Please refer to Figure 2 - Figure 4 As shown, the present invention also provides a foot health detection system based on a flexible capacitive sensor, comprising a data acquisition module 1, a data transmission module 2 and a computer module 3: the data acquisition module 1 collects the plantar pressure data of the subject in a free state through an intelligent insole with a flexible capacitive sensor, and transmits the collected plantar pressure data to the computer module 3 through the data transmission module 2; the data transmission module 2 can be a WIFI module, which can communicate with a host computer, and the data transmission module 2 is used to transmit the plantar pressure data collected by the data acquisition module 1 to the computer module 3; the computer module 3 is used to process, divide, train and denormalize the original plantar pressure data to obtain a model threshold;
[0113] Among them, the computer module 3 includes a processing module 31, a division module 32, a first training module 33, a second training module 34, and a denormalization processing module 35: the processing module 31 is used to perform data preprocessing, feature extraction and normalization processing on the original plantar pressure data, and establish a two-dimensional vector data set; the division module 32 is used to divide the data set into a training set and a test set in proportion; the first training module 33 trains the data in the training set through the SVR-LSTM neural network model to obtain model parameters; the second training module 34 trains the model parameters through the BP neural network optimized based on the improved Salp Insinuate algorithm to obtain plantar pressure prediction results; the denormalization processing module 35 obtains the model threshold by denormalizing the plantar pressure prediction results.
[0114] In this embodiment, the plantar detection system based on flexible capacitive sensors can more comprehensively process plantar pressure data, including its nonlinear characteristics, time series characteristics, and prediction tasks that require fine adjustment, thereby improving the accuracy and reliability of plantar pressure detection and providing an efficient and accurate solution for plantar health detection.
[0115] The working principle of the present invention is as follows: the foot sole detection system based on the flexible capacitive sensor is composed of a data acquisition module 1, a data transmission module 2 and a computer module 3, and the computer module 3 includes a processing module 31, a division module 32, a first training module 33, a second training module 34, and a denormalization processing module 35; when the foot sole pressure of the subject is detected by the foot sole detection method based on the flexible capacitive sensor, the subject first steps on the smart insole with the flexible capacitive sensor (such as Figure 1 As shown, the smart insole with a flexible capacitive sensor has nine collection areas. When the plantar pressure of the subject is collected, nine plantar pressure characteristic areas can be collected, including the first metatarsal, the first metatarsal, the second metatarsal, the third metatarsal, the fourth metatarsal, the fifth metatarsal, the midfoot, the inner side of the heel, the outer side of the heel and the back side of the heel). The smart insole with a flexible capacitive sensor collects the plantar pressure data of the subject in a free state, and transmits the collected plantar pressure data to a computer via WIFI; after the computer obtains the raw plantar pressure data, the processing module 31 performs data preprocessing, feature extraction and normalization on the raw plantar pressure data, and establishes a two-dimensional vector data set with the processed data value and the real data value; the division module 32 then divides the data set into two parts according to the proportion The first training module 33 trains the data in the training set through the SVR-LSTM neural network model to obtain the model parameters; the second training module 34 trains the model parameters through the BP neural network optimized based on the improved Salp Intestine algorithm to obtain the plantar pressure prediction results; finally, the denormalization processing module 35 denormalizes the plantar pressure prediction results and outputs the model threshold; the plantar detection system based on the flexible capacitive sensor of the present invention can more comprehensively process the plantar pressure data, including its nonlinear characteristics, time series characteristics and prediction tasks that need fine adjustment, thereby improving the accuracy and reliability of plantar pressure detection and providing an efficient and accurate solution for plantar health detection;
[0116] The present invention can better adapt to the foot shape of the testee by using an intelligent insole with a flexible capacitive sensor, provide a more comfortable detection experience, and further ensure that the plantar pressure data collected during the whole walking process of the testee can be closer to the actual situation, making the acquired data more accurate and objective; at the same time, in the present invention, the SVR model can process the complex pattern of plantar pressure data during the training process, and provide more accurate input for the subsequent LSTM model and BP neural network training based on the improved salp algorithm optimization. The LSTM model processes the dynamic characteristics of plantar pressure changing with time, and provides time series insights for the prediction of plantar pressure. The BP neural network optimized based on the improved salp algorithm optimizes network parameters by simulating the foraging behavior of salps, thereby improving the learning efficiency and prediction accuracy of the network.
[0117] Various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiment, 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, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logical function on a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0118] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A foot health detection method based on a flexible capacitive sensor, characterized in that: The following steps are involved: S1. Collect the plantar pressure data of the subject in a free state through a smart insole with a flexible capacitive sensor, and transmit the collected plantar pressure data to a computer via WIFI; S2, the computer obtains the original plantar pressure data and performs data preprocessing, feature extraction and normalization, and establishes a two-dimensional vector data set with the processed data value and the real data value; S3, divide the data set into training set and test set in proportion; S4, constructing an SVR-LSTM neural network model, and inputting the data in the training set into the SVR-LSTM neural network model for the first training; S5, inputting the model parameters obtained from the first training into the BP neural network optimized based on the improved Salp Integrity Algorithm to perform the second training; S6. Denormalize the plantar pressure prediction result obtained by the second training to obtain a model threshold.
2. The foot health detection method based on flexible capacitive sensor according to claim 1 is characterized in that: In step S1, the smart insole with a flexible capacitive sensor has nine collection areas. When collecting the plantar pressure of the subject, nine plantar pressure characteristic areas can be collected, including the first metatarsal, the first metatarsal, the second metatarsal, the third metatarsal, the fourth metatarsal, the fifth metatarsal, the midfoot, the inner side of the heel, the outer side of the heel and the back of the heel.
3. The foot health detection method based on flexible capacitive sensor according to claim 1 is characterized in that: In step S2, the following steps are included: S21, preprocessing the acquired plantar pressure data, including using a filter to remove various noises existing in the plantar pressure data; S22, when extracting features from the acquired plantar pressure data, including time domain eigenvalues, frequency domain eigenvalues, and nonlinear dynamics eigenvalues; S23, normalize the processed data, and the calculation formula is as follows: In the formula, X' is the normalized value of the original data, X min and X max They are the minimum and maximum values in the original data respectively. After normalization, the original data are mapped to the interval [0, 1].
4. The foot health detection method based on flexible capacitive sensor according to claim 1 is characterized in that: In step S3, the data set is a two-dimensional vector (X i ,Y i ), where X i is the processed data, Y i is the original data, that is, the true value of the data.
5. The foot health detection method based on flexible capacitive sensor according to claim 4 is characterized in that: In step S4, the following steps are included: S41. Build the SVR model and LSTM model, and the SVR model function is as follows: In the formula, K(x i ,x j ) is the kernel function, a i is the ith Lagrangian coefficient above the upper edge of the isolation zone, is the i-th Lagrangian coefficient below the upper edge of the isolation zone; S42, during the first training, the two-dimensional vector (X i ,Y i ) as training samples, and use SVR model and LSTM model for training respectively. The obtained threshold is recorded as SVR(X i ) and LSTM(X i ); S43, then the two-dimensional vector (SVR (X i ), LSTM(X i )) and the corresponding true value Y i The parameters are input as model parameters into the BP neural network optimized based on the improved salp algorithm, and the BP neural network optimized based on the improved salp algorithm is trained.
6. The method for detecting foot health based on a flexible capacitive sensor according to claim 5, characterized in that: In step S41, the LSTM model includes a memory unit, an input gate, an output gate and a forget gate, the input gate and the output gate are used to receive, output and correct parameters, the forget gate is used to selectively clear specific correction parameters, and the memory unit is used to record and save the current state of the neuron.
7. The foot health detection method based on flexible capacitive sensor according to claim 1 is characterized in that: In step S5, the BP neural network optimized based on the improved salp algorithm is to use the weights and thresholds of the BP neural network optimized based on the improved salp algorithm.
8. The foot health detection method based on flexible capacitive sensor according to claim 1 is characterized in that: In step S6, the calculation formula for the denormalization process is as follows: X”=X’(X max -X min )+X min ; In the formula, X' is the normalized data, and X" is the result after denormalization.
9. A foot health detection system based on a flexible capacitive sensor, characterized in that: include: A data acquisition module (1), wherein the data acquisition module (1) acquires plantar pressure data of a subject in a free state through an intelligent insole with a flexible capacitive sensor, and transmits the acquired plantar pressure data to a computer module (3) through a data transmission module (2); A data transmission module (2), the data transmission module (2) being used to transmit the plantar pressure data collected by the data collection module (1) to the computer module (3); A computer module (3) is used to process, divide, train and denormalize raw plantar pressure data to obtain a model threshold.
10. The foot health detection system based on flexible capacitive sensor according to claim 9, characterized in that: The computer module (3) comprises a processing module (31), a division module (32), a first training module (33), a second training module (34), and a denormalization processing module (35): The processing module (31) is used to perform data preprocessing, feature extraction and normalization processing on the original plantar pressure data, and establish a two-dimensional vector data set; The partitioning module (32) is used to divide the data set into a training set and a test set in proportion; The first training module (33) trains the data in the training set through the SVR-LSTM neural network model to obtain model parameters; The second training module (34) trains the model parameters by using a BP neural network optimized based on an improved Salp Integrity algorithm to obtain a plantar pressure prediction result; The denormalization processing module (35) performs denormalization processing on the plantar pressure prediction result to obtain a model threshold.
Citation Information
Cited By
Mangrove forest carbon sink dynamic monitoring method and system based on multi-source remote sensing fusion
CN120430513A
