Stress state assessment method and system based on Markov transition field of electrodermal signal

Abstract

The present disclosure provides a stress state assessment method and system based on the Markov transition field of skin electrodermal signals, which relates to the technical field of physiological signal analysis, including: obtaining skin electrodermal signals of subjects; constructing a Markov transition field, and using the Markov transition field to graphically transform the three-domain sequence of skin electrodermal signals into a grayscale image; coarsening the grayscale image, extracting the texture features of the processed grayscale image, and obtaining the graph features of the Markov transition field; constructing a feature set according to the feature selection result, and inputting the feature set into a classification model of a support vector machine (SVM) to obtain an assessment result of the stress state; wherein, the process of constructing the Markov transition field includes: determining the number of quantile intervals suitable for the skin electrodermal signals, dividing the skin electrodermal signals using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

Description

Technical Field

[0001] The present disclosure relates to the technical field of physiological signal analysis, and in particular to a method and system for assessing stress status based on a Markov transition field of an electrical skin signal. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The pace of life and work in modern society is getting faster and faster, social competition is rapidly intensifying, and people's mental health problems are prominent. Long-term psychological stress causes disorders in people's psychological and physiological functions, especially increases the risk of certain pathological diseases, and may even lead to sudden death. Therefore, early identification and assessment of psychological stress is very necessary. Among the current psychological stress assessment methods, questionnaire surveys are lagging and highly subjective; the information contained in biochemical markers is single, the measurement steps are complicated, and they are not universal. In comparison, physiological signals are more objective in assessing stress, easy to obtain, time-saving, non-invasive, and can realize real-time detection of stress. Among them, the skin electrode signal is more suitable for psychological stress monitoring in daily life because of its simple acquisition process and short analysis time.

[0004] In the current methods of using skin electrical signal to monitor psychological stress, the features extracted from the skin electrical signal in existing studies are relatively independent, and the effectiveness of the features is unclear. Although the effect of identifying multi-task stress based on skin electrical signal analysis is acceptable, its effect on identifying single-task stress needs to be improved. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a stress state assessment method and system based on the Markov transition field of skin electrodermal signals. The Markov transition field is used to convert the skin electrodermal signal time series, frequency domain series and time-frequency domain series into two-dimensional images respectively. After coarse-graining the three images, the local binary pattern features of the images are extracted, and the multi-domain features of the skin electrodermal signals are combined to construct a feature set to improve the recognition accuracy.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions:

[0007] The stress state assessment method based on the Markov transition field of the electrodermal signal includes:

[0008] Acquiring an electrical skin signal of a subject and preprocessing the electrical skin signal;

[0009] The preprocessed electrodermal signal is processed to obtain its time series, frequency domain series and time-frequency domain series;

[0010] Construct a Markov transition field, use the Markov transition field to graphically transform the three-domain sequence of the skin electrical signal, convert the color image into a grayscale image, and obtain the Markov image of the three domains;

[0011] Using mean pooling to perform coarse-graining processing on the three grayscale images, extracting texture features of the processed grayscale images, and obtaining graph features of the Markov transition field;

[0012] The mutual information, random forest and feature recursive elimination methods are used to select features of the graph. A feature set is constructed based on the feature selection results. The feature set is input into the classification model of the support vector machine (SVM) to obtain the evaluation results of the pressure state.

[0013] Among them, the process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the three-domain sequence of the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

[0014] According to some embodiments, the present disclosure adopts the following technical solutions:

[0015] The stress state assessment system based on the Markov transition field of electrodermal signals includes:

[0016] A signal preprocessing module, used to obtain the subject's electrical skin signal and preprocess the electrical skin signal;

[0017] A sequence extraction module is used to obtain the time series, frequency domain series and time-frequency domain series of the electrodermal signal;

[0018] A Markov transition field construction module is used to construct a Markov transition field for skin electrical signals;

[0019] A signal graphics module is used to graphically transform the three-domain sequence of the electrocutaneous signal using the Markov transition field, convert the color image into a grayscale image, and obtain the Markov image of the three domains;

[0020] Image processing module, using mean pooling to perform coarse-grained processing on grayscale images;

[0021] An image feature extraction module is used to extract texture features of the processed grayscale image and obtain the graph features of the Markov transition field;

[0022] The feature selection and classification module is used to select features of the graph using mutual information, random forest and feature recursive elimination methods, construct a feature set based on the feature selection results, and input the feature set into the classification model of the support vector machine (SVM) to obtain the evaluation results of the pressure state;

[0023] Among them, the process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the time domain sequence, frequency domain sequence and frequency domain sequence of the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

[0024] According to some embodiments, the present disclosure adopts the following technical solutions:

[0025] A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the pressure state assessment method based on the Markov transition field of the electrodermal signal is implemented.

[0026] According to some embodiments, the present disclosure adopts the following technical solutions:

[0027] An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the stress state assessment method based on the Markov transition field of the electrodermal signal.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention uses the Markov model to process the time domain sequence, frequency domain sequence and time-frequency domain sequence of the electrodermal signal. On the basis of the traditional Markov model that only processes the time series, this method introduces the frequency domain sequence and the time-frequency domain sequence, thereby extracting more useful feature information and providing richer information for psychological stress detection.

[0030] The stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention performs coarse-grained processing on the grayscale image through mean pooling, thereby reducing redundant details in the image and highlighting useful feature information. This improvement not only helps the subsequent extraction of image features, making the extracted image features more concise and effective, but also improves the overall effect of feature extraction.

[0031] The stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention uses the grayscale value of the central pixel as the threshold and compares it with the grayscale values ​​of its adjacent pixels to obtain a corresponding binary value to represent the local texture features of the image, which is used to measure the uniformity, homogeneity and similarity of the image grayscale distribution, and quantify the local change boundary of the image grayscale distribution. LBP (Local Binary Patterns) is used to extract the graph features of the Markov transition field, and a total of 10 LBP features are obtained, which fully mine the potential information of the electrodermal signal and overcome the disadvantage that the traditional features based on the electrodermal signal are not effective in identifying a single stress task.

[0032] The stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention constructs a Markov transition field, describes the relationship between sequences with the help of the single-step and multi-step transfer probabilities defined in the Markov chain, and provides more additional information for the electrodermal signal to detect the psychological stress state by extracting the LBP features of the image.

[0033] The stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention uses the Markov transition field to graph the electrodermal signal to obtain a two-dimensional image, thereby realizing the visualization of a one-dimensional time series; the graphical analysis method can also be applied to other physiological signals to detect stress or other states. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0035] Figure 1 is a method flow chart of an embodiment of the present disclosure;

[0036] Figure 2 A flowchart of constructing a Markov transition field according to an embodiment of the present disclosure;

[0037] Figure 3 This is an example diagram of LBP coding calculation according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0038] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0039] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0040] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0041] Example 1

[0042] In one embodiment of the present disclosure, a stress state assessment method based on the Markov transition field of the skin electrical signal is provided. The Markov transition field is used to convert the time series, frequency domain series and time-frequency domain series of the skin electrical signal into a two-dimensional image. After coarse-grained processing, the local binary pattern features of the image are extracted. The multi-domain features of the skin electrical signal are combined to construct a feature set to improve the recognition accuracy. At the same time, a new analysis direction is provided for other physiological signals. The method includes:

[0043] Step 1: obtaining the subject's electrical skin signal and preprocessing the electrical skin signal;

[0044] Step 2: Process the preprocessed skin electrical signal to calculate the time series, frequency domain series and time-frequency domain series of the skin electrical signal;

[0045] Step 3: Construct a Markov transition field, use the Markov transition field to graphically transform the three-domain sequence of the skin electrical signal, convert the color image into a grayscale image, and obtain the Markov image of the three domains;

[0046] Step 4: Perform coarse-graining processing on the three grayscale images to reduce redundant details in the images and highlight useful information;

[0047] Step 5: Extract the texture features of the processed grayscale image and obtain the graph features of the Markov transition field;

[0048] Step 6: Use mutual information, random forest and feature recursive elimination methods to select features of the graph, build a feature set based on the feature selection results, input the feature set into the support vector machine (SVM) classification model, and obtain the evaluation results of the pressure state;

[0049] The process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

[0050] As an embodiment, the specific implementation process of a stress state assessment method based on the Markov transition field of the electrodermal signal disclosed in the present invention is as follows:

[0051] Step 1: Collect the subject's electrical skin signal and pre-process the collected electrical skin signal;

[0052] Specifically, the subject's electrical skin signal is collected and filtered: a Butterworth filter is used for bandpass filtering with a frequency range of 0.01 to 2 Hz to remove noise and interference components, thereby obtaining a relatively smooth and ideal electrical skin signal for subsequent processing.

[0053] During training, two types of skin electrical signals are collected in the calm state and the task stress state. To evaluate the stress state using skin electrical signals, it is necessary to be able to distinguish between the stress state and the calm state, so it is necessary to collect skin electrical signals in the stress state. The processing of the two states is the same, both are Markov graphing and then extracting LBP features, but the LBP feature values ​​extracted from the two states are different. The purpose of collecting the calm state and the stress state is to prove that the method of Markov graphing and then extracting features can effectively detect the stress state.

[0054] Step 2: Perform Fourier transform on the electrodermal signal to obtain a frequency domain sequence, and perform wavelet transform on the electrodermal signal to obtain a two-dimensional time-frequency diagram. For the time-frequency diagram, add a time window in the time domain direction, calculate the mean, median, maximum and minimum values, and standard deviation of the frequency domain within the time window, and obtain 5 different time-frequency domain sequences by continuously sliding the time window.

[0055] Step 3: Use the Markov transition field to graphically transform the three-domain sequence of the skin electrical signal and convert the color image into a grayscale image.

[0056] Specifically, it is necessary to construct a Markov transition field first. To construct a Markov transition field, it is necessary to first determine the number of quantile intervals Q suitable for the skin electrical signal. Q determines the order of the Markov transition matrix W. Then, the skin electrical signal is divided using the quantile intervals to form different quantile interval types. The Markov transition matrix is ​​constructed according to the intervals. The values ​​of the elements in the Markov transition matrix are counted and normalized. The normalized values ​​are arranged in chronological order to expand and generate the Markov transition field. The specific steps are as follows:

[0057] 1) Quantile interval division, including: converting the acquired skin electrical signal into a time series, discretizing the continuous time series frequency domain sequence and time-frequency domain sequence using the quantile method, making multiple quantile intervals have the same area under the Gaussian curve through multiple quantiles, forming an ordered list of quantile points of the quantile interval, and the area of ​​the curve between the quantile points and the time axis approximately obeys the distribution.

[0058] Specifically, for a given skin electrical signal, let its time series be X = {x 1 ,x 2 ,...,x n}, the frequency domain sequence is Y = {y 1 ,y 2 ,...,y n}, the time-frequency domain sequence is Z = {z 1 ,z 2 ,...,z n}. Use the quantile method to discretize the continuous sequences X and Y. Taking the time series as an example, Q-1 quantiles are used to make Q quantile intervals have the same area under the Gaussian curve. Therefore, the quantile points of the quantile interval form an ordered list Q = {q 1 ,q 2 ,...,q Q}, and the area S of the curve between the quantiles and the time axis approximately follows the following distribution:

[0059]

[0060] Among them, q j is the jth quantile interval, q j+1 represents the j+1th quantile interval.

[0061] 2) Construct a Markov transition matrix. According to the division of quantile intervals, map each value in the time, frequency domain and time-frequency domain sequences to the corresponding quantile intervals. The transition probability between quantile intervals is calculated by the transition frequency between each quantile interval in the three sequences. The relationship between sequences is described according to the single-step and multi-step transition probabilities defined in the Markov chain. Combined with the multiple discretely distributed quantile intervals obtained after the three-domain amplitude discretization of the electrodermal signal, the continuous time, frequency domain and time-frequency domain sequences are converted into quantile sequences. The transition of the quantile sequence at each moment is statistically analyzed to obtain the transition probability. The transition probability of the Markov chain is arranged in the set direction to construct the weighted adjacency matrix. After normalizing the weighted adjacency matrix, the Markov transition matrix is ​​obtained.

[0062] Specifically, based on the division of quantile intervals, taking time series as an example, each value x in the time series t is mapped to the corresponding quantile interval q i The transition probability between intervals can be calculated by the transition frequency between intervals in the sequence, and the relationship between sequences can be described by the single-step and multi-step transition probabilities defined in the Markov chain, as follows:

[0063]

[0064] Among them, P i,jis the multi-step transition probability, indicating that it is currently in the quantile interval q i The elements of the next moment transition to the quantile interval q in multiple steps j Combining the Q discrete distribution quantile intervals obtained after discretization of the signal's time domain amplitude, the continuous time series can be converted into a quantile sequence, and the transition probability can be obtained by counting the changes in the quantile sequence at each moment. t The transition probability of the Markov chain is arranged in the increasing direction to construct the weighted adjacency matrix W, w ij Indicates that in the interval q i A point in the next time step is in the interval q j The probability of using ∑ j w ij =1 After normalizing the weight adjacency matrix W, the resulting matrix W is the Markov transition matrix, as shown below, where P is the probability:

[0065]

[0066] 3) Constructing a Markov transition field, including expanding the transition probability matrix into the Markov transition field by considering the time position information. The Markov transition field uses the transition probability to encode the transition probability of multi-span time series, representing the transition probability between two points of a set time interval and the transition probability between two points of a frequency interval. The Markov transition field directly expresses the probability information of the Markov transition matrix.

[0067] Specifically, the Markov chain has the characteristic of no memory. The probability distribution of the current state is only determined by its previous state and has nothing to do with other states in the time series and frequency domain sequence, that is, it satisfies the formula,

[0068]

[0069] The Markov transition matrix constructed based on the Markov chain is also memoryless. This method abandons the time distribution of X and the time dependence of the time step. Directly using the Markov transition matrix will lead to the neglect of too much time series X information. The Markov transition field M arranges the transition probabilities in a regular time sequence, which can overcome the disadvantage of time information loss. The same is true for frequency domain and time-frequency domain sequences. Taking time series as an example, the definition of the Markov transition field is as follows:

[0070]

[0071] In the Markov transition field M, the quantile intervals of the values ​​at time steps i and j are q i and q j , M ij Represents the interval q i to q jFor example, let the time series X = {x 1 ,x 2 ,...,x n} is quantized to 8 quantile intervals Q = {q 1 ,q 2 ,...,q 8}, assuming the value x at time step t t The quantile interval is q 6 , the value x at time step 2 2 The quantile interval is q 3 , at this time q 6 and q 3 Corresponding to the above q i and q j , then in the Markov transition field M, M t2 Represents the quantile interval q 6 to q 3 The transition probability corresponds to the Markov transition matrix W 63 The transition probability matrix W is extended to the Markov transition field M by considering the temporal position information. The Markov transition field M uses the transition probability M ij Encodes the transition probability of multiple span time series, M ij||i-j|＝k represents the transition probability between two points with a time interval of k. For example, M ij||j-i|＝1 represents the probability of a one-step transition along the time axis, M 15 Represents the transition probability of the interval between two points with a time interval of 4 along the time axis. The main diagonal M ii It is quite special in that it captures the probability of each quantile transferring to itself, which is the self-transition probability. Therefore, the Markov transition field M can directly express the probability information of the Markov transition matrix W.

[0072] Furthermore, the electrodermal signal is graphically visualized. In the Markov transition field M, M ij It can be regarded as a pixel point, whose value range is [0,1]. The following formula can be used to scale the value of each element in the matrix to 0-255 so that it corresponds to the pixel value in the image, thereby obtaining a two-dimensional image and realizing the visualization of the one-dimensional time series.

[0073] I(i,j)=int(255M ij )

[0074] Where I(i,j) is the pixel value of the i-th row and j-th column of the image; int(·) is the rounding function.

[0075] Step 4: Coarse-grain the image to reduce redundant detail features and highlight useful features.

[0076] Specifically, the purpose of image coarsening is to reduce the details of the image by reducing the resolution of the image. This can be achieved by merging the color values ​​or grayscale values ​​of multiple adjacent pixels into a new pixel value. Therefore, the mean pooling method is used to process the image. The processing steps include: first, dividing the image into blocks of fixed size, then taking the average value of the pixel values ​​in each block, and finally replacing all the pixel values ​​in the block with the average value to achieve coarse-grained processing.

[0077] Step 5: Extract the texture features of the processed image and obtain the graph features of the Markov transition field. This includes: using the grayscale value of the central pixel as the threshold, comparing it with the grayscale values ​​of its adjacent pixels, and obtaining the corresponding binary value to represent the local texture features of the image. The pixel to the left of the central pixel is defined as the starting point, and the value is taken in a counterclockwise direction to obtain the LBP code of the image in the area. Each pixel has a corresponding LBP code, and the graph feature map is obtained, which is called the LBP feature map.

[0078] Specifically, the core principle of LBP is to use the gray value of the central pixel as the threshold and compare it with the gray values ​​of its adjacent pixels to obtain the corresponding binary value to represent the local texture features of the image, which is used to measure the uniformity, homogeneity, and similarity of the image gray distribution and quantify the local change boundary of the image gray distribution. c ,y c ), LBP calculation can be expressed as:

[0079]

[0080] Among them, g c is the center pixel (x c ,y c ) gray value, g p is the grayscale value of the neighborhood pixel, P is the number of neighborhood pixels on the circle with R as radius, and sgn(·) is the sign function.

[0081] Figure 3The principle of the feature map is as follows: LBP is very simple in principle, and its operation is equivalent to locally binarizing the difference between the grayscale value of the central pixel and the grayscale value of its adjacent pixels. When the grayscale value of the neighborhood pixel is greater than the grayscale value of the central pixel, the value of the pixel is recorded as a binary number 1, otherwise it is recorded as 0, and the resulting binary string is LBP. If P is 8 and R is 1, the pixel to the left of the central pixel is defined as the starting point, and the value is taken in a counterclockwise direction to obtain the LBP result of the image in the area, which is the LBP code, which is between 0 and 255 in decimal. Similarly, if all pixels in the image are processed using LBP, each pixel has a corresponding LBP code, and the feature map obtained in this way is called an LBP feature map.

[0082] The equivalent LBP means that the binary number corresponding to the pattern contains at most two transitions from 1 to 0 or from 0 to 1 (pattern type U≤2). P,R arrive (superscript u2 indicates equivalent mode) mapping, the original 2 p The dimensionality of the modes is reduced to p*(p-1)+3, which reduces the amount of calculation while retaining most of the effective information in the original image. In order to achieve rotation invariance, the local rotation invariance mode is defined as follows:

[0083]

[0084] In the above formula, from LBP P,R arrive (the superscript riu2 represents the rotationally invariant equivalent mode U≤2) and produces p+2 different LBP eigenvalues.

[0085] LBP is used to extract the graph features of the Markov transition field, p is set to 8, and a total of 10 LBP features are obtained.

[0086] Step 6: Use mutual information, random forest and feature recursive elimination methods to select features of the graph, build a feature set based on the feature selection results, input the feature set into the support vector machine (SVM) classification model, and obtain the evaluation results of the pressure state;

[0087] Specifically, firstly, a support vector machine (SVM) classification model was constructed, and the existing network structure was used as the classification model. During training, three methods, namely mutual information, random forest and feature recursive elimination method, were used to select features of resting and two stress states. The feature importance ranking of LBP features and joint feature sets was given to facilitate the construction of the optimal feature subset.

[0088] Example 2

[0089] In one embodiment of the present disclosure, a stress state assessment system based on a Markov transition field of an electrodermal signal is provided, comprising:

[0090] A signal preprocessing module, used to obtain the subject's electrical skin signal and preprocess the electrical skin signal;

[0091] A sequence extraction module is used to obtain the time series, frequency domain series and time-frequency domain series of the electrodermal signal;

[0092] A Markov transition field construction module is used to construct a Markov transition field for the preprocessed electrocutaneous signal;

[0093] A signal graphics module is used to perform a graphical conversion on the pre-processed electrocutaneous signal using a Markov transition field, and convert a color image into a grayscale image;

[0094] The image processing module uses mean pooling to coarse-grain grayscale images, remove unnecessary details in the image, and highlight useful features;

[0095] An image feature extraction module, used to extract texture features of the grayscale image and obtain graph features of the Markov transition field;

[0096] The feature selection and classification module is used to select features of the graph using mutual information, random forest and feature recursive elimination methods, construct a feature set based on the feature selection results, and input the feature set into the classification model of the support vector machine (SVM) to obtain the evaluation results of the pressure state;

[0097] The process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

[0098] Example 3

[0099] In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the pressure state assessment method based on the Markov transition field of the electrodermal signal is implemented.

[0100] Example 4

[0101] In one embodiment of the present disclosure, an electronic device is provided, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the stress state assessment method based on the Markov transition field of the electrodermal signal.

[0102] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0104] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A stress state assessment method based on the Markov transition field of skin electrical signals, characterized in that: include: Acquiring an electrical skin signal of a subject and preprocessing the electrical skin signal; Extract the time, frequency and time-frequency domain sequences of the preprocessed skin electrical signal, use the three-domain sequences to construct the Markov transition field, use the Markov transition field to perform graphical transformation on the preprocessed skin electrical signal, transform the color image into a grayscale image, and obtain the three-domain Markov image; Extract the texture features of the processed grayscale image and obtain the graph features of the Markov transition field, including: using the grayscale value of the central pixel as a threshold, comparing it with the grayscale values ​​of its adjacent pixels, thereby obtaining a corresponding binary value to represent the local texture features of the image, defining the pixel to the left of the central pixel as the starting point, taking values ​​in a counterclockwise direction, and obtaining the LBP code of the local image of the region, each pixel has a corresponding LBP code, and obtaining a graph feature map, which is called an LBP feature map; Use mean pooling to coarse-grain the grayscale image, extract the texture features of the processed grayscale image, and obtain the graph features of the Markov transition field; During training, three methods, namely mutual information, random forest and feature recursive elimination, were used to select features of resting state and two stress states, and the feature importance ranking of LBP features and joint feature sets was given to construct the optimal feature subset. The subset was input into the classification model of support vector machine (SVM) to obtain the evaluation results of stress state, in which the existing network structure was used as the classification model. The process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

2. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 1, characterized in that: The acquired skin electrical signal is band-pass filtered, and the frequency range is adjusted to remove noise and interference components to obtain a smooth and ideal skin electrical signal.

3. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 1, characterized in that: The EP signal is divided by using quantile intervals, including: converting the acquired EP signal into a time series, extracting its frequency domain and time-frequency domain series, discretizing the continuous time series and frequency domain series by using the quantile method, making multiple quantile intervals have the same area under the Gaussian curve through multiple quantiles, forming an ordered list of the quantile points of the quantile intervals, and the curves between the quantile points approximately follow the distribution of the areas of the time axis and the frequency axis.

4. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 1, characterized in that: Construct a Markov transition matrix, including: According to the division of quantile intervals, each value in the three-domain sequence is mapped to the corresponding quantile interval. The transition probability between quantile intervals is calculated through the transition frequency between each quantile interval in the time, frequency domain and time-frequency domain sequences. The relationship between the time and frequency domain sequences is described according to the single-step and multi-step transition probabilities defined in the Markov chain.

5. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 4 is characterized in that: Combining the multiple discretely distributed quantile intervals obtained by discretizing the time domain, frequency domain and time-frequency domain amplitude of the electrodermal signal, the continuous three-domain sequence is converted into a quantile sequence, and the changes of the quantile sequence at each moment are statistically analyzed to obtain the transition probability. The transition probability of the Markov chain is arranged according to the set direction to construct a weighted adjacency matrix. After normalizing the weighted adjacency matrix, the Markov transition matrix is ​​obtained.

6. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 1, characterized in that: Constructing a Markov transition field includes: expanding the transition probability matrix into the Markov transition field by considering the time, frequency domain and time-frequency domain position information; the Markov transition field uses the transition probability to encode the transition probability of a multi-span time series, indicating the transition probability between two points of a set time interval; the Markov transition field directly expresses the probability information of the Markov transition matrix.

7. The stress state assessment method based on the Markov transition field of the electrodermal signal according to claim 1, characterized in that: The grayscale image is coarse-grained, including: dividing the image into blocks of fixed size, then taking the average value of the pixel values ​​in each block, and finally replacing all the pixel values ​​in the block with the average value to achieve coarse-grained processing.

8. A stress state assessment system based on the Markov transition field of electrodermal signals, characterized in that: Specifically implementing the stress state assessment method based on the Markov transition field of the electrodermal signal as described in any one of claims 1 to 7 includes: A signal preprocessing module, used to obtain the subject's electrical skin signal and preprocess the electrical skin signal; A sequence extraction module is used to obtain the time, frequency and time-frequency domain sequences of the electrodermal signal; A Markov transition field construction module is used to construct a Markov transition field for skin electrical signals; The signal graphics module is used to use the Markov transition field to perform a graphical transformation on the three-domain sequence of the electrodermal signal and convert the color image into a grayscale image; An image processing module for coarse-graining grayscale images using mean pooling; An image feature extraction module is used to extract texture features of the processed grayscale image and obtain the graph features of the Markov transition field; The feature selection and classification module is used to select the features of the resting state and the two stress states using three methods: mutual information, random forest and feature recursive elimination method during training, give the feature importance ranking of LBP features and joint feature sets, construct the optimal feature subset, input the subset into the classification model of support vector machine (SVM), and obtain the evaluation result of the stress state, wherein the existing network structure is used as the classification model; The process of constructing a Markov transition field includes: determining the number of quantile intervals suitable for the electrodermal signal, dividing the electrodermal signal using the quantile intervals, and forming different quantile interval categories; constructing a Markov transition matrix according to the quantile interval categories, and expanding the Markov transition matrix to generate a Markov transition field.

9. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the stress state assessment method based on the Markov transition field of the electrodermal signal as described in any one of claims 1-7.

Images