An Adaptive Denoising Method for PPG Signals Based on Wavelet Transform and Autoencoder

Abstract

The present invention relates to the technical field of signal adaptive noise reduction, and provides a PPG signal adaptive noise reduction method based on wavelet transform and autoencoder; this method combines adaptive variational mode decomposition, stationary wavelet denoising, multi-dimensional consistency optimization, synchronous update technology and Lagrange multiplier constraint adjustment technology to remove high-frequency noise from the PPG signal, and can better adapt to the individual differences of drivers and the complexity of signals; the total loss function of the autoencoder is optimized by using contrastive loss, weighted reconstruction error and sparsity regularization, which enhances the sensitivity and judgment ability for abnormal health states; this method can accurately restore and evaluate the PPG signal in multiple dimensions, provides reliable technical support for the continuous monitoring of drivers' health states, has high stability and scalability, and can provide important guarantees for vehicle safe driving and driver health management.

Description

Technical Field

[0001] The present invention relates to the technical field of signal adaptive noise reduction, and in particular to an adaptive noise reduction method for PPG signals based on wavelet transform and autoencoder. Background Art

[0002] The adaptive noise reduction method for PPG signals (photoplethysmogram signal, the PPG signal is a biomedical signal that uses the optical principle to detect the change in blood volume under the skin) based on wavelet transform and autoencoder is a technology that combines wavelet transform and deep learning autoencoder to process the noise in the PPG signal. The wavelet transform decomposes the signal in the time and frequency domains and can effectively separate the noise components; PPG is a non-invasive biosignal measurement technology that uses optical methods to detect the change in peripheral blood volume of the human body. The PPG signal reflects the change in blood volume in the human blood vessels with the cardiac pumping cycle. The PPG signal usually contains a DC component and an AC component and is widely used in health monitoring, including heart rate monitoring, blood oxygen saturation measurement, and blood pressure estimation.

[0003] In driver health monitoring, the PPG signal is an important basis for evaluating the driver's physiological state, and the signal quality of the PPG signal is crucial for the accuracy of evaluating the driver's health status; however, due to the complexity of signal noise in the dynamic driving environment, the traditional wavelet transform noise reduction method is difficult to handle the multi-dimensional characteristics of the PPG signal. The traditional wavelet transform noise reduction method cannot adaptively balance the signal and noise components, resulting in signal distortion and poor noise reduction effect, and cannot provide high-quality data support for evaluating the driver's health status; at the same time, the traditional autoencoder is difficult to comprehensively capture the deep features in multi-dimensional physiological signals, thus affecting the accuracy of anomaly detection; therefore, there is an urgent need for a technical solution that can adaptively remove the noise in multi-dimensional PPG signals and accurately evaluate the driver's health status, providing a strong guarantee for the health management and early warning of drivers. Summary of the Invention

[0004] In the present invention, a denoising algorithm of MAVMD-SWT (Adaptive Variational Mode Decomposition and Stationary Wavelet Transform Denoising Algorithm) is constructed to denoise multi-dimensional PPG signals. The MAVMD-SWT denoising algorithm first preliminarily decomposes the multi-dimensional PPG signals by using adaptive variational mode decomposition, then uses the synchronous update technology and the Lagrange multiplier constraint adjustment technology to ensure the balance and consistency of noise and signal components in the multi-dimensional space, and finally performs denoising through stationary wavelet transform, providing high-quality data support for subsequent health status assessment; to further improve the data quality after signal denoising, an enhanced sparse autoencoder is designed to optimize the total loss function of the autoencoder by combining contrast loss, weighted reconstruction error, weight decay term and sparsity regularization term, so as to realize deep feature learning and health status classification of the denoised data, and be used to accurately judge the health status of the driver; the invention not only improves the effectiveness of the data, but also provides reliable safety guarantee in the fields of driving safety and health monitoring.

[0005] The present invention provides a PPG signal adaptive denoising method based on wavelet transform and autoencoder, and the method includes the following steps:

[0006] Step T1: Data acquisition: Collect the heart rate data, heart rate variability data, blood oxygen saturation data and pulse waveform data of the driver, and fuse them to obtain multi-dimensional comprehensive health data;

[0007] Step T2: Data preprocessing: Synchronize the signal of the multi-dimensional comprehensive health data, remove the DC offset, filter and standardize it to obtain preprocessed multi-dimensional comprehensive health data;

[0008] Step T3: Data denoising: Combine adaptive variational mode decomposition, stationary wavelet denoising, multi-dimensional consistency optimization, synchronous update technology and Lagrange multiplier constraint adjustment technology to obtain the MAVMD-SWT denoising algorithm, and denoise the preprocessed multi-dimensional comprehensive health data through the MAVMD-SWT denoising algorithm to obtain denoised multi-dimensional comprehensive health data; the MAVMD-SWT denoising algorithm specifically includes a decomposition layer initialization unit, a multi-dimensional VMD decomposition unit (multi-dimensional variational mode decomposition unit), a key value calculation unit, a key value judgment unit, a multi-dimensional stationary wavelet denoising unit and a multi-dimensional signal reconstruction unit;

[0009] Step T4: Anomaly detection: Optimize the total loss function of the autoencoder by combining contrast loss, weighted reconstruction error, weight decay term and sparsity regularization term to obtain an enhanced sparse autoencoder, input the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder, perform driver anomaly detection, and judge the health status of the driver.

[0010] Further, in step T3, the steps executed by the decomposition layer initialization unit specifically include setting the initial decomposition layer number of VMD to 2.

[0011] Further, in step T3, the steps performed by the multi-dimensional VMD decomposition unit specifically include performing VMD decomposition on the preprocessed multi-dimensional comprehensive health data through multiple channels, and decomposing to obtain multi-dimensional IMF components (multi-dimensional intrinsic mode function components) according to the initial decomposition layer number of VMD.

[0012] Further, in step T3, the steps performed by the key value calculation unit specifically include calculating the kurtosis and correlation coefficient of each dimension for the multi-dimensional IMF components respectively, and taking the product of the kurtosis and the correlation coefficient as the key value.

[0013] Further, in step T3, the steps performed by the key value judgment unit specifically include checking the minimum key value of the IMF components of each dimension in the multi-dimensional IMF components. If the minimum key value ≤ 1, mark the IMF component as a high-noise component and enter stationary wavelet denoising; if the minimum key value > 1, increase the decomposition layer number and re-perform multi-dimensional VMD decomposition until multi-dimensional high-noise IMF components and multi-dimensional noise-free IMF components are marked; obtain a multi-dimensional consistency optimization algorithm through synchronous update technology and Lagrange multiplier constraint adjustment technology; use the multi-dimensional consistency optimization algorithm to further select and synchronize the multi-dimensional high-noise IMF components to obtain optimized multi-dimensional high-noise IMF components.

[0014] Further, in step T3, the steps performed by the multi-dimensional stationary wavelet denoising unit specifically include selecting a wavelet basis for each dimension of the optimized multi-dimensional high-noise IMF components, and applying stationary wavelet transform to remove high-frequency noise to obtain denoised multi-dimensional IMF components.

[0015] Further, in step T3, the steps performed by the multi-dimensional signal reconstruction unit specifically include recombining the denoised multi-dimensional IMF components and the multi-dimensional noise-free IMF components, and performing multi-dimensional reconstruction to form denoised multi-dimensional comprehensive health data.

[0016] Further, in the key value judgment unit, the process of using the multi-dimensional consistency optimization algorithm to further select and synchronize the multi-dimensional high-noise IMF components to obtain optimized multi-dimensional high-noise IMF components specifically includes the following steps:

[0017] Step S1: Construct an objective function: Construct a multi-dimensional joint optimization objective function, and the formula used is as follows:

[0018] ;

[0019] Among them, represents the IMF component index, represents the th IMF component, denotes the center frequency of the th IMF component; denotes the multi-dimensional joint optimization objective function, denotes the total number of dimensions, denotes the dimension index, denotes the weight coefficient of the th dimension, denotes time derivative of, denotes the imaginary unit, denotes the complex exponential function with the center frequency, denotes the square of the norm;

[0020] Step S2: Update of IMF components: Based on the multi-dimensional joint optimization objective function, synchronously update the multi-dimensional high-noise IMF components to gradually reduce the noise in each dimension. The formula used is as follows:

[0021] ;

[0022] where and both denote the IMF component index, denotes the frequency variable, denotes the index of the iteration number, denotes the th IMF component's spectrum update value in the th iteration, denotes the spectrum of the original signal, denotes the representation of the th IMF component in the frequency domain, denotes the sum of the spectra of other IMF components except the th IMF component, denotes the Lagrange multiplier, denotes the smoothing parameter;

[0023] Step S3: Update of Lagrange multiplier: After each synchronous update, introduce the Lagrange multiplier for constraint adjustment to ensure the frequency consistency of the multi-dimensional high-noise IMF components. Iteratively update the parameters of the multi-dimensional high-noise IMF components and the Lagrange multiplier, and finally output the optimized multi-dimensional high-noise IMF components. The formula used is as follows:

[0024] ;

[0025] where denotes the index of the iteration number, denotes the updated value of the Lagrange multiplier in the th iteration, denotes the The Lagrange multiplier of the next iteration represents the update step coefficient represents the th IMF component at the update value at the th iteration,

[0026] Furthermore, in step T4, the process of inputting the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder for driver anomaly detection and judging the driver's health status specifically includes the following steps:

[0027] Step T41: Data input: The enhanced sparse autoencoder includes an encoding layer, a hidden layer, and a decoding layer. Initialize the weights of the enhanced sparse autoencoder and input the denoised multi-dimensional comprehensive health data into the encoding layer;

[0028] Step T42: Forward propagation encoding: Convert the denoised multi-dimensional comprehensive health data into a low-dimensional sparse representation through the encoding layer, and set the sparsity parameter to control the activity level of the hidden layer neurons;

[0029] Step T43: Calculate the sparsity loss: Calculate the sparsity regularization term through the KL divergence to measure the deviation between the actual average activation value of the hidden layer neurons and the sparsity parameter. The formula used is as follows:

[0030] ;

[0031] where, represents the sparsity regularization term, represents the neuron index of the hidden layer, represents the total number of neurons in the hidden layer, represents the summation symbol, represents the sparsity parameter, represents the th actual average activation value of the hidden layer neurons, represents the KL divergence, represents the first part of the KL divergence formula to measure the deviation between the th actual average activation value of the hidden layer neurons and the sparsity parameter in the activated state, represents the second part in the KL divergence formula to measure the non-activated state deviation of the th hidden layer neurons;

[0032] Step T44: Backpropagation optimization: Introduce the contrast loss, weighted reconstruction error, and weight decay term, and combine the sparsity regularization term to construct the total loss function. Minimize the total loss function through backpropagation to adjust the weights of the enhanced sparse autoencoder. The formula used is as follows:

[0033] ;

[0034] wherein, represents the index of the sample, represents the total number of samples, represents the index of the feature, represents the total number of features, represents the summation over samples, represents the summation over each feature dimension in each sample, represents the -th weight of the feature, represents the -th sample's -th original input value of the feature, represents the -th sample's -th reconstructed input value of the feature, represents the squared error on the -th feature dimension of samples; represents the weight of the sparsity regularization term, and represent the neuron index, represents the connection weight value from the -th neuron to the -th neuron, represents the square of the weight; represents the weight coefficient of the contrast loss, represents the contrast loss;

[0035] Step T45: Health status assessment: According to the weights of the enhanced sparse autoencoder, the low-dimensional sparse representation is restored through the decoding layer to obtain the reconstructed data. The reconstruction error is calculated based on the reconstructed data, and the driver's health status is judged according to the reconstruction error.

[0036] Adopting the above scheme, the beneficial effects achieved by the present invention are as follows:

[0037] The present invention realizes the adaptive noise reduction processing of multi-dimensional PPG signals by constructing the MAVMD-SWT denoising algorithm, effectively improving the quality and reliability of driver health monitoring data. First, the algorithm uses the adaptive variational mode decomposition technology to preliminarily decompose the multi-dimensional signals, and then through the synchronous update and Lagrange multiplier constraint adjustment technology, it ensures the balance and consistency of each signal component in the multi-dimensional space. Finally, it further removes noise by combining the stationary wavelet transform. This adaptive denoising process overcomes the limitations of traditional filtering and fixed mode decomposition methods in dealing with complex noises in the dynamic driving environment, significantly improving the accuracy of signal denoising, providing high-quality input data support for subsequent health status assessment, and ensuring clear and reliable physiological data in different environments.

[0038] Meanwhile, the present invention further improves the feature learning and the accuracy of health status assessment of the denoised data by designing an enhanced sparse autoencoder. The enhanced sparse autoencoder optimizes the loss function by combining the contrast loss, weighted reconstruction error, weight decay term, and sparsity regularization term, effectively extracts the deep features of the PPG signal data, and realizes the classification and anomaly detection of the driver's health status based on this. Compared with the traditional autoencoder, the enhanced sparse autoencoder of the present invention can accurately identify fatigue, low blood oxygen, and abnormal heart rate, effectively enhancing the anomaly warning ability of the health monitoring system and improving the reliability and accuracy of health assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic flow chart of a PPG signal adaptive noise reduction method based on wavelet transform and autoencoder proposed by the present invention;

[0040] Figure 2 It is a schematic flow chart of the multi-dimensional consistency optimization algorithm in the key value judgment unit proposed by the present invention;

[0041] Figure 3 It is a schematic flow chart of the enhanced sparse autoencoder in step T4 proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0043] Embodiment 1. According to Figure 1 , the present invention provides a PPG signal adaptive noise reduction method based on wavelet transform and autoencoder, and this method includes the following steps:

[0044] Step T1: Data acquisition: Collect the heart rate data, heart rate variability data, blood oxygen saturation data, and pulse waveform data of the driver, and fuse them to obtain multi-dimensional comprehensive health data;

[0045] Step T2: Data preprocessing: Synchronize the signals of the multi-dimensional comprehensive health data, remove the DC offset, filter, and standardize it to obtain preprocessed multi-dimensional comprehensive health data;

[0046] Step T3: Data denoising: Combine the adaptive variational mode decomposition, stationary wavelet denoising, multi-dimensional consistency optimization, synchronous update technology, and Lagrange multiplier constraint adjustment technology to obtain the MAVMD-SWT denoising algorithm. Denoise the preprocessed multi-dimensional comprehensive health data through the MAVMD-SWT denoising algorithm to obtain denoised multi-dimensional comprehensive health data; The MAVMD-SWT denoising algorithm specifically includes a decomposition layer initialization unit, a multi-dimensional VMD decomposition unit, a key value calculation unit, a key value judgment unit, a multi-dimensional stationary wavelet denoising unit, and a multi-dimensional signal reconstruction unit;

[0047] Step T4: Anomaly detection: Combine the contrast loss, weighted reconstruction error, weight decay term, and sparsity regularization term to optimize the total loss function of the autoencoder to obtain an enhanced sparse autoencoder. Input the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder for driver anomaly detection to judge the health status of the driver.

[0048] Embodiment 2: Based on the above embodiment, in Step T3, the steps performed by the decomposition layer initialization unit specifically include setting the initial VMD decomposition layer number to 2.

[0049] Embodiment 3: Based on the above embodiment, in Step T3, the steps performed by the multi-dimensional VMD decomposition unit specifically include performing VMD decomposition on the preprocessed multi-dimensional comprehensive health data through multiple channels, and decomposing it into multi-dimensional IMF components according to the initial VMD decomposition layer number.

[0050] Embodiment 4: Based on the above embodiment, in Step T3, the steps performed by the key value calculation unit specifically include calculating the kurtosis and correlation coefficient of each dimension for the multi-dimensional IMF components respectively, and taking the product of the kurtosis and the correlation coefficient as the key value.

[0051] Example 5. This example is based on the above example. In step T3, the steps performed by the key value judgment unit specifically include checking the minimum key value of the IMF components in each dimension of the multi-dimensional IMF components. If the minimum key value ≤ 1, mark the IMF component as a high-noise component and enter stationary wavelet denoising; if the minimum key value > 1, increase the decomposition level and perform multi-dimensional VMD decomposition again until multi-dimensional high-noise IMF components and multi-dimensional noise-free IMF components are marked; obtain a multi-dimensional consistency optimization algorithm through synchronous update technology and Lagrange multiplier constraint adjustment technology; use the multi-dimensional consistency optimization algorithm to further select and synchronize the multi-dimensional high-noise IMF components to obtain optimized multi-dimensional high-noise IMF components.

[0052] Example 6. This example is based on the above example. In step T3, the steps performed by the multi-dimensional stationary wavelet denoising unit specifically include selecting wavelet bases for each dimension of the optimized multi-dimensional high-noise IMF components, applying stationary wavelet transform to remove high-frequency noise, and obtaining denoised multi-dimensional IMF components.

[0053] Example 7. This example is based on the above example. In step T3, the steps performed by the multi-dimensional signal reconstruction unit specifically include recombining the denoised multi-dimensional IMF components and the multi-dimensional noise-free IMF components, performing multi-dimensional reconstruction, and forming denoised multi-dimensional comprehensive health data.

[0054] Example 8. According to Figure 2 , this example is based on the above example. In the key value judgment unit, the process of using the multi-dimensional consistency optimization algorithm to further select and synchronize the multi-dimensional high-noise IMF components to obtain optimized multi-dimensional high-noise IMF components specifically includes the following steps:

[0055] Step S1: Construct an objective function: Construct a multi-dimensional joint optimization objective function, and the formula used is as follows:

[0056] ;

[0057] Among them, represents the IMF component index, represents the th IMF component, represents the th central frequency of the IMF component; represents the multi-dimensional joint optimization objective function, represents the total number of dimensions, represents the dimension index, represents the th weight coefficient of the dimension, represents the total number of IMF components, represents time The derivative of represents the imaginary unit, represents a complex exponential function with a center frequency, represents the square of the norm;

[0058] Step S2: IMF component update: Based on the multi-dimensional joint optimization objective function, the multi-dimensional high-noise IMF components are synchronously updated to gradually reduce the noise in each dimension. The formula used is as follows:

[0059] ;

[0060] in, and Both represent IMF component indexes. represents a frequency variable, The index representing the number of iterations, Indicates The IMF component is The spectrum update value in the iteration, represents the spectrum of the original signal, Indicates The representation of the IMF components in the frequency domain is: Indicates that except The sum of the spectra of the other IMF components except the IMF components, represents the Lagrange multiplier, represents the smoothing parameter;

[0061] Step S3: Lagrange multiplier update: After each synchronization update, Lagrange multipliers are introduced to perform constraint adjustment to ensure the frequency consistency of the multidimensional high-noise IMF component, iteratively update the parameters and Lagrange multipliers of the multidimensional high-noise IMF component, and finally output the optimized multidimensional high-noise IMF component. The formula used is as follows:

[0062] ;

[0063] in, The index representing the number of iterations, Indicates The updated value of the Lagrange multiplier at the iteration, Indicates The Lagrange multiplier of the iteration, represents the update step coefficient, Indicates The IMF component is The updated value at the iteration, represents the sum of the spectra of the updated IMF components;

[0064] This algorithm first uses the adaptive variational mode decomposition technique to preliminarily decompose the multi-dimensional signal, and then through the synchronous update and Lagrange multiplier constraint adjustment technique, it ensures the balance and consistency of each signal component in the multi-dimensional space. Finally, it further removes noise by combining with the stationary wavelet transform. This adaptive denoising process overcomes the limitations of traditional filtering and fixed mode decomposition methods in dealing with complex noises in the dynamic driving environment, significantly improves the accuracy of signal denoising, provides high-quality input data support for subsequent health status assessment, and ensures clear and reliable physiological data can be obtained in different environments.

[0065] Embodiment Nine. According to Figure 3 , based on the above embodiment, in step T4, the process of inputting the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder for driver anomaly detection and judging the driver's health status specifically includes the following steps:

[0066] Step T41: Data input: The enhanced sparse autoencoder includes an encoding layer, a hidden layer, and a decoding layer. Initialize the weights of the enhanced sparse autoencoder and input the denoised multi-dimensional comprehensive health data into the encoding layer;

[0067] Step T42: Forward propagation encoding: Convert the denoised multi-dimensional comprehensive health data into a low-dimensional sparse representation through the encoding layer, and set the sparsity parameter to control the activity degree of the neurons in the hidden layer;

[0068] Step T43: Calculate the sparsity loss: Calculate the sparsity regularization term through the KL divergence, measure the deviation between the actual average activation value of the neurons in the hidden layer and the sparsity parameter, and the formula used is as follows:

[0069] ;

[0070] Where represents the sparsity regularization term, represents the neuron index of the hidden layer, represents the total number of neurons in the hidden layer, represents the summation symbol, represents the sparsity parameter, represents the th actual average activation value of the neurons in the hidden layer, represents the KL divergence, represents the first part of the KL divergence formula to measure the deviation between the actual average activation value of the th neuron in the hidden layer and the sparsity parameter in the activated state, represents the second part in the KL divergence formula to measure the non-activated state deviation of the th neuron in the hidden layer;

[0071] Step T44: Backpropagation optimization: Introduce contrastive loss, weighted reconstruction error, and weight decay terms, and construct a total loss function in combination with a sparsity regularization term. Minimize the total loss function through backpropagation to adjust the weights of the enhanced sparse autoencoder. The formula used is as follows:

[0072] ;

[0073] Among them, represents the index of the sample, represents the total number of samples, represents the index of the feature, represents the total number of features, represents the summation over samples, represents the summation over feature dimensions in each sample, represents the weight of the th feature, represents the th sample's th feature's original input value, represents the th sample's th feature's reconstructed input value, represents the th sample's th feature dimension's squared error; represents the weight of the sparsity regularization term, represents the control coefficient of the weight decay term; and represent neuron indices, represents from the th neuron to the th neuron's connection weight value, represents the square of the weight; represents the weight coefficient of the contrastive loss, represents the contrastive loss;

[0074] Step T45: Health status assessment: According to the weights of the enhanced sparse autoencoder, restore the low-dimensional sparse representation through the decoding layer to obtain reconstructed data. Calculate the reconstruction error based on the reconstructed data, and judge the driver's health status based on the reconstruction error;

[0075] The present invention further improves the accuracy of feature learning and health status assessment of the denoised data by designing an enhanced sparse autoencoder; the enhanced sparse autoencoder optimizes the loss function by combining contrast loss, weighted reconstruction error, weight decay term, and sparsity regularization term, effectively extracts the deep features of the PPG signal data, and realizes the classification and anomaly detection of the driver's health status based on this; compared with the traditional autoencoder, the enhanced sparse autoencoder of the present invention can accurately identify fatigue, low blood oxygen, and abnormal heart rate, effectively enhances the anomaly warning ability of the health monitoring system, and improves the reliability and accuracy of health assessment.

[0076] The above describes the present invention and its implementation manners. Such a description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto; in general, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments without creative work without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. A PPG signal adaptive denoising method based on wavelet transform and autoencoder, characterized in that: The following steps are involved: Step T1: Data collection: Collect the driver's heart rate data, heart rate variability data, blood oxygen saturation data and pulse waveform data, and fuse them to obtain multi-dimensional comprehensive health data; Step T2: Data preprocessing: performing signal synchronization, DC offset removal, filtering and standardization on the multi-dimensional comprehensive health data to obtain pre-processed multi-dimensional comprehensive health data; Step T3: Data denoising: Combining adaptive variational mode decomposition, stationary wavelet denoising, multidimensional consistency optimization, synchronous update technology and Lagrange multiplier constraint adjustment technology, the MAVMD-SWT denoising algorithm is obtained, and the pre-processed multidimensional comprehensive health data is denoised by the MAVMD-SWT denoising algorithm to obtain denoised multidimensional comprehensive health data; the MAVMD-SWT denoising algorithm specifically includes a decomposition layer initialization unit, a multidimensional VMD decomposition unit, a key value calculation unit, a key value judgment unit, a multidimensional stationary wavelet denoising unit and a multidimensional signal reconstruction unit; the multidimensional VMD decomposition unit decomposes to obtain multidimensional IMF components; Step T4: Anomaly detection: Combine contrast loss, weighted reconstruction error, weight decay term and sparsity regularization term to optimize the total loss function of the autoencoder to obtain an enhanced sparse autoencoder. Input the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder to perform driver anomaly detection and judge the driver's health status. In step T3, the key value calculation unit performs the following steps: calculating the kurtosis and correlation coefficient of each dimension of the multidimensional IMF component, and taking the product of the kurtosis and the correlation coefficient as the key value; In step T3, the key value judgment unit executes the following steps specifically: checking the minimum key value of the IMF component of each dimension in the multidimensional IMF component; if the minimum key value is ≤1, marking the IMF component as a high-noise component and entering stationary wavelet denoising; if the minimum key value is >1, increasing the number of decomposition layers and re-performing multidimensional VMD decomposition until a multidimensional high-noise IMF component and a multidimensional noise-free IMF component are marked; obtaining a multidimensional consistency optimization algorithm through synchronous update technology and Lagrange multiplier constraint adjustment technology; and using the multidimensional consistency optimization algorithm to further select and synchronize the multidimensional high-noise IMF components to obtain optimized multidimensional high-noise IMF components.

2. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 1, characterized in that: In step T3, the decomposition layer initialization unit executes the steps including setting the initial VMD decomposition layer to 2.

3. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 2, characterized in that: In step T3, the multidimensional VMD decomposition unit executes the steps of performing VMD decomposition on the pre-processed multidimensional comprehensive health data through multiple channels, and decomposing the pre-processed multidimensional comprehensive health data into multidimensional IMF components according to the number of initial VMD decomposition layers.

4. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 3, characterized in that: In step T3, the multidimensional stationary wavelet denoising unit performs the following steps: selecting a wavelet basis for each dimension of the optimized multidimensional high-noise IMF component, applying stationary wavelet transform to remove high-frequency noise, and obtaining a denoised multidimensional IMF component.

5. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 4, characterized in that: In step T3, the steps performed by the multidimensional signal reconstruction unit specifically include recombining the denoised multidimensional IMF components and the multidimensional noise-free IMF components, performing multidimensional reconstruction, and forming denoised multidimensional comprehensive health data.

6. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 5, characterized in that: In the key value judgment unit, the multidimensional high-noise IMF component is further selected and synchronized using the multidimensional consistency optimization algorithm to obtain a process of optimizing the multidimensional high-noise IMF component, which specifically includes the following steps: Step S1: construct objective function: construct multi-dimensional joint optimization objective function; Step S2: IMF component update: based on the multi-dimensional joint optimization objective function, the multi-dimensional high-noise IMF components are synchronously updated to gradually reduce the noise in each dimension; Step S3: Lagrange multiplier update: After each synchronization update, Lagrange multipliers are introduced to perform constraint adjustment to ensure the frequency consistency of the multidimensional high-noise IMF component, and the parameters and Lagrange multipliers of the multidimensional high-noise IMF component are iteratively updated to finally output the optimized multidimensional high-noise IMF component.

7. The method for adaptive denoising of PPG signals based on wavelet transform and autoencoder according to claim 1, characterized in that: In step T4, the process of inputting the denoised multi-dimensional comprehensive health data into the enhanced sparse autoencoder to perform driver abnormality detection and determine the driver's health status specifically includes the following steps: Step T41: Data input: The enhanced sparse autoencoder includes an encoding layer, a hidden layer, and a decoding layer. The weights of the enhanced sparse autoencoder are initialized, and the denoised multi-dimensional comprehensive health data is input into the encoding layer; Step T42: forward propagation encoding: convert the denoised multi-dimensional comprehensive health data into a low-dimensional sparse representation through the encoding layer, and set the sparsity parameters to control the activity of the hidden layer neurons; Step T43: Calculate the sparsity loss: Calculate the sparsity regularization term through KL divergence to measure the deviation between the actual average activation value of the hidden layer neurons and the sparsity parameter; Step T44: Back propagation optimization: introduce contrast loss, weighted reconstruction error and weight decay terms, and construct a total loss function in combination with the sparsity regularization term. Minimize the total loss function through back propagation and adjust the weights of the enhanced sparse autoencoder. Step T45: Health status assessment: According to the weights of the enhanced sparse autoencoder, the low-dimensional sparse representation is restored through the decoding layer to obtain reconstructed data, the reconstruction error is calculated based on the reconstructed data, and the driver's health status is judged based on the reconstruction error.

Images