Attention analysis method, system and device based on micro-action detection and storage medium

Abstract

The invention discloses an attention analysis method, system and device based on micro-motion detection and a storage medium, and the method comprises the steps: obtaining the micro-motion data of a target object, and carrying out the preprocessing of the micro-motion data, and obtaining the preprocessing data; wherein the micro-action data comprises pressure data and inertial data, and the pre-processing data comprises pressure pre-processing data and inertial pre-processing data; calculating a characteristic index value based on the pressure pre-processing data and the inertia pre-processing data; inputting the feature index value into a pre-trained deep learning model to obtain an attention state classification result of the target object; based on the attention state classification result, attention trend analysis is carried out, an attention change curve is constructed, and duration and conversion frequencies of different attention states and mutation points of the attention change curve are obtained. The attention state of the target object is analyzed through the micro-action data of the target object, and real-time and accurate analysis of the attention state of the target object is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent teaching technology, and specifically relates to an attention analysis method, system, device and storage medium based on micro-motion detection. Background Art

[0002] Traditional methods of assessing student attention mainly rely on teacher observation or self-reporting, which have problems such as strong subjectivity and poor real-time performance. With the rapid development of educational technology, a variety of innovative methods have emerged in the field of student attention assessment.

[0003] Most of the existing innovative technologies for attention assessment tend to use various testing equipment to objectively assess students' attention. Most of these technologies have problems such as expensive equipment, complex operation, or great interference with the learning environment. At the same time, existing testing equipment cannot capture students' subtle changes in movement, making it difficult to achieve real-time detection, resulting in inaccurate and in-time assessment results of students' attention status. On this basis, it is even more difficult to combine test data to effectively analyze the long-term trend of students' attention. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide an attention analysis method, system, device and storage medium based on micro-motion detection, which can analyze the attention state of the target object by analyzing its tiny movement changes, and realize real-time and accurate analysis of the attention state of the target object.

[0005] The present invention provides the following technical solutions:

[0006] In a first aspect, a method for attention analysis based on micro-motion detection is provided, comprising: acquiring micro-motion data of a target object, and preprocessing the micro-motion data to obtain preprocessed data; wherein the micro-motion data comprises pressure data and inertia data, and the preprocessed data comprises pressure preprocessed data and inertia preprocessed data; calculating characteristic index values ​​based on the pressure preprocessed data and the inertia preprocessed data; inputting the characteristic index values ​​into a pre-trained deep learning model to obtain an attention state classification result of the target object; performing attention trend analysis based on the attention state classification result and constructing an attention change curve to obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve.

[0007] As a preferred technical solution of the present invention, the pressure data is collected by multiple pressure sensors arranged on the seat of the target object; the inertial data includes acceleration data and angular velocity data, which are respectively collected by a three-axis accelerometer and a three-axis gyroscope arranged on the seat of the target object.

[0008] As a preferred technical solution of the present invention, the preprocessing of micro-motion data includes:

[0009] The pressure data is subjected to median filtering, which is expressed as:

[0010] ;

[0011] in, represents the filtered pressure data, represents median filtering, represents pressure data, n represents the nth sampling point of pressure data in discrete time, k represents the offset of the nth sampling point, N represents the length of the sliding window, and N is an odd number;

[0012] The inertial data is subjected to a first-order low-pass filter, which is expressed as:

[0013] ;

[0014] in, represents the filtered inertial data, represents the inertial data, m represents the mth sampling point of the inertial data at discrete time, , represents the cutoff frequency of the filter, represents the sampling period, represents the time constant;

[0015] The filtered pressure data is normalized to [0, 1], expressed as:

[0016] ;

[0017] Where P represents the pressure data of a single pressure sensor at any sampling point in the filtered pressure data. Indicates the minimum value of the corresponding pressure data P, Indicates the maximum value of the corresponding pressure data P, represents the normalized pressure data;

[0018] The filtered inertial data is normalized to [0, 1], expressed as:

[0019] ;

[0020] Where I represents the acceleration data or angular velocity data at any sampling point in the filtered inertial data, Indicates the minimum value of the corresponding acceleration data or angular velocity data I, Indicates the maximum value of the corresponding acceleration data or angular velocity data I, Indicates normalized acceleration data or angular velocity data;

[0021] The normalized pressure data and inertia data are respectively divided into time windows according to set time, and pressure preprocessing data and inertia preprocessing data are obtained respectively; wherein the inertia preprocessing data includes acceleration preprocessing data or angular velocity preprocessing data.

[0022] As a preferred technical solution of the present invention, the calculation of characteristic index values ​​based on pressure preprocessing data and inertia preprocessing data includes:

[0023] Calculate the center of gravity of pressure preprocessing data , expressed as:

[0024] ;

[0025] ;

[0026] in, The horizontal coordinate represents the center of gravity. The ordinate represents the center of gravity. represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and j-th column, M represents the total number of rows, and J represents the total number of columns;

[0027] Calculate the rate of change of center of gravity , expressed as:

[0028] ;

[0029] ;

[0030] in, represents the rate of change of the horizontal coordinate of the center of gravity position, It represents the rate of change of the vertical coordinate of the center of gravity position, represents the horizontal coordinate of the center of gravity at time t, The ordinate represents the center of gravity at time t;

[0031] when When it is greater than the set value, a sitting posture change is recorded; the number of sitting posture changes within a period of time is counted to obtain the sitting posture change frequency;

[0032] Calculate the pressure distribution entropy value H, expressed as:

[0033] ;

[0034] in, represents the probability distribution of pressure preprocessing data, , represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and the j-th column;

[0035] The pressure preprocessing data of each time window Flatten into a pressure data vector, center the pressure data vectors of all time windows to make their mean zero, and calculate the covariance matrix of the pressure data vector, which is expressed as:

[0036] ;

[0037] Where C represents the covariance matrix, E represents the total number of time windows, Indicates The pressure data vector of a time window, T represents transposition;

[0038] Perform eigenvalue decomposition on the covariance matrix C and obtain the eigenvalue and the corresponding eigenvector ; Select the one with the largest eigenvalue The eigenvectors are used as principal components, and the pressure preprocessing data are projected onto the principal components to obtain dimensional PCA eigenvector of pressure distribution.

[0039] As a preferred technical solution of the present invention, the calculation of characteristic index values ​​based on pressure preprocessing data and inertia preprocessing data also includes:

[0040] The mean, variance, kurtosis and skewness of the acceleration preprocessed data or the angular velocity preprocessed data are calculated and expressed as:

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] in, Indicates that any axis is Acceleration preprocessing data or angular velocity preprocessing data of a time window, , Represents the x-axis, y-axis and z-axis respectively. represents the mean, E represents the total number of time windows, represents the variance, represents the kurtosis, represents skewness;

[0046] Perform fast Fourier transform on the acceleration preprocessing data and the angular velocity preprocessing data to obtain corresponding frequency spectra; calculate the energy and frequency center of the acceleration preprocessing data or the angular velocity preprocessing data, which are respectively expressed as:

[0047] ;

[0048] ;

[0049] in, Indicates energy, represents the frequency center, represents the frequency spectrum of acceleration preprocessing data or angular velocity preprocessing data, f represents the fth frequency component in the frequency spectrum, and F represents the total number of frequency components;

[0050] Calculating the attitude angle and attitude angle change rate through acceleration preprocessing data and angular velocity preprocessing data; wherein the attitude angle includes pitch angle, roll angle and yaw angle;

[0051] Calculate the amplitude of motion, expressed as:

[0052] ;

[0053] in, Indicates the range of motion, Represents the acceleration preprocessing data of the x, y, and z axes respectively.

[0054] As a preferred technical solution of the present invention, the step of inputting the characteristic index value into a pre-trained deep learning model to obtain the attention state classification result of the target object includes:

[0055] The characteristic index values ​​in each time window are combined into a characteristic vector, and the characteristic vectors in each time window are arranged in chronological order to form a characteristic matrix, which is expressed as:

[0056] ;

[0057] Where X represents the feature matrix, Represents the sth characteristic index value in the Eth time window;

[0058] The feature matrix X is input into the pre-trained deep learning model. In each time window e, the deep learning model receives the feature vector of the current time window. and the hidden state of the previous time window , and output the hidden state of the current time window ;

[0059] The hidden states of all time windows are input into the deep learning model, and after the hidden states of all time windows are integrated, they are input into the fully connected layer to obtain the corresponding probability values ​​of each attention state, which is expressed as:

[0060] ;

[0061] in, represents the probability value of the cth attention state, represents the cth element of the fully connected layer output, and C represents the total number of attention state categories;

[0062] Based on the corresponding probability value of each attention state, the probability distribution of the attention state is obtained, which is expressed as:

[0063] ;

[0064] Where p represents the probability distribution of attention state;

[0065] The attention state with the largest probability value in the probability distribution p is taken as the attention state classification result.

[0066] As a preferred technical solution of the present invention, the attention trend analysis is performed based on the attention state classification result and the attention change curve is constructed to obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve, including:

[0067] The attention state classification results in each time window form an attention state sequence, and the duration of each attention state classification result appearing continuously is counted to obtain the duration of each attention state; the number of times each attention state classification result changes within a set time is counted to obtain the conversion frequency of each attention state;

[0068] The exponential smoothing method is used to smooth the attention state sequence to obtain the attention change curve, which is expressed as:

[0069] ;

[0070] in, represents the attention state in the e-th time window, represents the smoothed attention state, represents the smoothing factor;

[0071] The Pettitt test algorithm is used to detect the mutation points in the attention change curve;

[0072] Fourier transform was used to analyze whether there was a periodic pattern in the attention change curve.

[0073] In a second aspect, an attention analysis system based on micro-motion detection is provided, comprising: a data preprocessing module, used to obtain micro-motion data of a target object, and preprocess the micro-motion data to obtain preprocessed data; wherein the micro-motion data includes pressure data and inertia data, and the preprocessed data includes pressure preprocessed data and inertia preprocessed data;

[0074] A feature calculation module, used for calculating a feature index value based on the pressure preprocessing data and the inertia preprocessing data;

[0075] A classification module, used to input the characteristic index value into a pre-trained deep learning model to obtain a classification result of the attention state of the target object;

[0076] The attention analysis module is used to perform attention trend analysis and construct an attention change curve based on the attention state classification result, and obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve.

[0077] In a third aspect, there is provided an attention analysis device based on micro-motion detection, comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method described in the first aspect.

[0078] According to a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] The present invention provides an attention analysis method based on micro-motion detection, which analyzes the attention state of a target object through its micro-motion data. The micro-motion detection equipment has low cost, simple operation, and little interference to the environment, and can detect tiny movement changes of the target object in real time. By analyzing the micro-motion data in real time, it is convenient for teachers to understand the attention state of the target object in a timely manner, and at the same time, it can realize the long-term trend analysis of the target object's attention, which is beneficial for teachers to optimize teaching strategies and improve teaching effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 is a flow chart of an attention analysis method based on micro-motion detection in an embodiment of the present invention;

[0082] Figure 2 is a graph showing changes in attention in an embodiment of the present invention. DETAILED DESCRIPTION

[0083] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0084] Example 1

[0085] This embodiment provides an attention analysis method based on micro-motion detection. Figure 1 As shown, the specific steps include:

[0086] Step 1: Acquire micro-motion data of the target object, and pre-process the micro-motion data to obtain pre-processed data.

[0087] Specifically, the micro-motion data includes pressure data and inertia data, and the pre-processed data includes pressure pre-processed data and inertia pre-processed data.

[0088] In this embodiment, the pressure data is collected by a plurality of pressure sensors arranged on the seat of the target object. A plurality of pressure sensors form a pressure sensor array, which is arranged on the seat cushion and backrest of the seat to form a 16×16 grid. Each pressure sensor collects pressure data at a frequency of 50 Hz.

[0089] In this embodiment, the inertial data includes acceleration data and angular velocity data, which are respectively collected by a three-axis accelerometer and a three-axis gyroscope arranged on the seat of the target object. The three-axis accelerometer and the three-axis gyroscope are installed above the seat back and collect the motion data of the upper body of the target object at a frequency of 100 Hz.

[0090] The micro-motion data is preprocessed as follows:

[0091] 1.1. Median filtering is performed on the pressure data, which is expressed as:

[0092] ;

[0093] in, represents the filtered pressure data, represents median filtering, represents pressure data, n represents the nth sampling point of pressure data in discrete time, k represents the offset of the nth sampling point, N represents the length value of the sliding window, and N is an odd number.

[0094] This step can effectively remove impulse noise and outliers in the pressure data and retain useful low-frequency information.

[0095] 1.2. Perform a first-order low-pass filter on the inertial data, expressed as:

[0096] ;

[0097] in, represents the filtered inertial data, represents the inertial data, m represents the mth sampling point of the inertial data at discrete time, , represents the cutoff frequency of the filter, represents the sampling period, Indicates the time constant. Adjust the time constant The filter cutoff frequency can be controlled to effectively remove high-frequency noise and jitter, while retaining low-frequency motion information related to the target object's attention.

[0098] 1.3. Normalize the filtered pressure data to [0, 1], expressed as:

[0099] ;

[0100] Where P represents the pressure data of a single pressure sensor at any sampling point in the filtered pressure data. Indicates the minimum value of the corresponding pressure data P, Indicates the maximum value of the corresponding pressure data P, Represents normalized pressure data.

[0101] 1.4. Normalize the filtered inertial data to [0, 1], expressed as:

[0102] ;

[0103] Where I represents the acceleration data or angular velocity data at any sampling point in the filtered inertial data, Indicates the minimum value of the corresponding acceleration data or angular velocity data I, Indicates the maximum value of the corresponding acceleration data or angular velocity data I, Indicates normalized acceleration data or angular velocity data.

[0104] Furthermore, I represents acceleration data or angular velocity data collected by a single-axis accelerometer or a single-axis gyroscope.

[0105] Steps 1.3 and 1.4 are to eliminate the dimensional differences between different sensors and improve the accuracy of analysis.

[0106] 1.5. The normalized pressure data and inertia data are divided into time windows according to the set time, and pressure pre-processing data and inertia pre-processing data are obtained respectively. The inertia pre-processing data includes acceleration pre-processing data or angular velocity pre-processing data.

[0107] In this embodiment, the segmentation is performed according to 5-second time windows, and each window overlaps 50%.

[0108] Step 2: Calculate characteristic index values ​​based on the pressure preprocessing data and the inertia preprocessing data.

[0109] (1) Calculating characteristic index values ​​based on the pressure preprocessing data, as follows:

[0110] (1.1) Calculate the center of gravity of the pressure preprocessing data , expressed as:

[0111] ;

[0112] ;

[0113] in, The horizontal coordinate represents the center of gravity. The ordinate represents the center of gravity. It represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and j-th column, M represents the total number of rows, and J represents the total number of columns.

[0114] (1.2) Calculate the rate of change of the center of gravity , expressed as:

[0115] ;

[0116] ;

[0117] in, represents the rate of change of the horizontal coordinate of the center of gravity position, It represents the rate of change of the vertical coordinate of the center of gravity position, represents the horizontal coordinate of the center of gravity at time t, Represents the vertical coordinate of the center of gravity at time t.

[0118] (1.3) When When it is greater than the set value, a sitting posture change is recorded; the number of sitting posture changes within a period of time is counted to obtain the sitting posture change frequency.

[0119] (1.4) Calculate the pressure distribution entropy value H, expressed as:

[0120] ;

[0121] in, represents the probability distribution of pressure preprocessing data, , Represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and j-th column.

[0122] (1.5) Preprocess the pressure data of each time window Flatten into a pressure data vector, center the pressure data vectors of all time windows to make their mean zero, and calculate the covariance matrix of the pressure data vector, which is expressed as:

[0123] ;

[0124] Where C represents the covariance matrix, E represents the total number of time windows, Indicates The pressure data vector of a time window, T represents the transpose.

[0125] Perform eigenvalue decomposition on the covariance matrix C and obtain the eigenvalue and the corresponding eigenvector ; Select the one with the largest eigenvalue The eigenvectors are used as principal components, and the pressure preprocessing data are projected onto the principal components to obtain dimensional pressure distribution PCA (Principal Component Analysis) eigenvector.

[0126] (2) Calculating characteristic index values ​​based on the inertial preprocessing data, as follows:

[0127] (2.1) The mean, variance, kurtosis and skewness of the acceleration preprocessed data or the angular velocity preprocessed data are calculated and expressed as:

[0128] ;

[0129] ;

[0130] ;

[0131] ;

[0132] in, Indicates that any axis is Acceleration preprocessing data or angular velocity preprocessing data of a time window, , Represents the x-axis, y-axis and z-axis respectively. represents the mean, E represents the total number of time windows, represents the variance, represents the kurtosis, Represents skewness.

[0133] (2.2) Performing fast Fourier transform on the acceleration preprocessed data and the angular velocity preprocessed data to obtain corresponding frequency spectra; calculating the energy and frequency center of the acceleration preprocessed data or the angular velocity preprocessed data, which are expressed as:

[0134] ;

[0135] ;

[0136] in, Indicates energy, represents the frequency center, represents the frequency spectrum of acceleration preprocessing data or angular velocity preprocessing data, f represents the fth frequency component in the frequency spectrum, and F represents the total number of frequency components.

[0137] (2.3) Calculating attitude angle and attitude angle change rate by using acceleration preprocessing data and angular velocity preprocessing data; wherein the attitude angle includes pitch angle, roll angle and yaw angle.

[0138] (2.4) Calculate the motion amplitude, expressed as:

[0139] ;

[0140] in, Indicates the range of motion, Represents the acceleration preprocessing data of the x, y, and z axes respectively.

[0141] Furthermore, the action amplitude is continuously greater than the threshold The length of time is taken as the duration of the action.

[0142] Step 3: Input the characteristic index value into the pre-trained deep learning model to obtain the attention state classification result of the target object. The details are as follows:

[0143] 3.1. The characteristic index values ​​in each time window are combined into a characteristic vector, and the characteristic vectors in each time window are arranged in chronological order to form a characteristic matrix, which is expressed as:

[0144] ;

[0145] Where X represents the feature matrix, Represents the sth feature index value in the Eth time window.

[0146] 3.2. Input the feature matrix X into the pre-trained deep learning model. In each time window e, the deep learning model receives the feature vector of the current time window and the hidden state of the previous time window , and output the hidden state of the current time window .

[0147] 3.3. Input the hidden states of all time windows into the deep learning model, integrate the hidden states of all time windows, and input them into the fully connected layer to obtain the corresponding probability values ​​of each attention state, expressed as:

[0148] ;

[0149] in, represents the probability value of the cth attention state, represents the cth element of the fully connected layer output, and C represents the total number of attention state categories.

[0150] 3.4. Based on the corresponding probability values ​​of each attention state, the probability distribution of the attention state is obtained, which is expressed as:

[0151] ;

[0152] Where p represents the probability distribution of attention state.

[0153] 3.5. The attention state with the largest probability value in the probability distribution p is taken as the attention state classification result.

[0154] In this embodiment, the attention state includes four levels: "high concentration", "normal", "mild distraction" and "severe distraction". The probability distribution of the attention state is p=[0.1, 0.6, 0.2, 0.1] (corresponding to "high concentration", "normal", "mild distraction" and "severe distraction" respectively), among which the probability value corresponding to "normal" is the highest, so the attention state in the current time window is determined to be "normal".

[0155] Step 4: Based on the attention state classification results, perform attention trend analysis and construct an attention change curve to obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve. The details are as follows:

[0156] (1) The attention state classification results within each time window form an attention state sequence. The duration of each attention state classification result is counted to obtain the duration of each attention state; the number of times each attention state classification result changes within the set time is counted to obtain the conversion frequency of each attention state. The duration and conversion frequency of each attention state reflect the concentration of the target object, so that teachers can understand the state of the target object in time and take corresponding measures to improve teaching effectiveness.

[0157] (2) The exponential smoothing method is used to smooth the attention state sequence to obtain the attention change curve, which is expressed as:

[0158] ;

[0159] in, represents the attention state in the e-th time window, represents the smoothed attention state, represents the smoothing factor;

[0160] The Pettitt test algorithm is used to detect the mutation points in the attention change curve. Figure 2 As shown, these mutation points correspond to moments in the learning process when attention suddenly decreases or increases.

[0161] Fourier transform is used to analyze whether there is a periodic pattern in the attention change curve. For example, after studying for a period of time, students' attention decreases and then recovers, forming a periodic fluctuation.

[0162] Example 2

[0163] This embodiment provides an attention analysis system based on micro-motion detection, including:

[0164] The data preprocessing module is used to obtain micro-motion data of the target object and preprocess the micro-motion data to obtain preprocessed data; wherein the micro-motion data includes pressure data and inertia data, and the preprocessed data includes pressure preprocessed data and inertia preprocessed data.

[0165] The feature calculation module is used to calculate the feature index value based on the pressure preprocessing data and the inertia preprocessing data.

[0166] The classification module is used to input the characteristic index value into a pre-trained deep learning model to obtain the attention state classification result of the target object.

[0167] The attention analysis module is used to perform attention trend analysis and construct an attention change curve based on the attention state classification result, and obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve.

[0168] Example 3

[0169] This embodiment provides an attention analysis device based on micro-motion detection, including a processor and a storage medium. The storage medium is used to store instructions. The processor is used to operate according to the instructions to execute the steps of the attention analysis method based on micro-motion detection described in Example 1.

[0170] Example 4

[0171] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the attention analysis method based on micro-motion detection described in Embodiment 1 are implemented.

[0172] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0173] The present application 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 application. 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.

[0174] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0175] 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.

[0176] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for attention analysis based on micro-motion detection, characterized in that: include: Acquire micro-motion data of the target object, and pre-process the micro-motion data to obtain pre-processed data; wherein the micro-motion data includes pressure data and inertia data, and the pre-processed data includes pressure pre-processed data and inertia pre-processed data; Calculating a characteristic index value based on the pressure preprocessing data and the inertia preprocessing data; Inputting the characteristic index value into a pre-trained deep learning model to obtain a classification result of the attention state of the target object; Based on the attention state classification result, attention trend analysis is performed and an attention change curve is constructed to obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve.

2. The attention analysis method based on micro-motion detection according to claim 1, characterized in that: The pressure data is collected by a plurality of pressure sensors arranged on the seat of the target object; the inertial data includes acceleration data and angular velocity data, which are collected by a three-axis accelerometer and a three-axis gyroscope arranged on the seat of the target object, respectively.

3. The attention analysis method based on micro-motion detection according to claim 2, characterized in that: The preprocessing of the micro-motion data includes: The pressure data is subjected to median filtering, which is expressed as: ； in, represents the filtered pressure data, represents median filtering, represents pressure data, n represents the nth sampling point of pressure data in discrete time, k represents the offset of the nth sampling point, N represents the length of the sliding window, and N is an odd number; The inertial data is subjected to a first-order low-pass filter, which is expressed as: ； in, represents the filtered inertial data, represents the inertial data, m represents the mth sampling point of the inertial data at discrete time, , represents the cutoff frequency of the filter, represents the sampling period, represents the time constant; The filtered pressure data is normalized to [0, 1], expressed as: ； Where P represents the pressure data of a single pressure sensor at any sampling point in the filtered pressure data. Indicates the minimum value of the corresponding pressure data P, Indicates the maximum value of the corresponding pressure data P, Represents normalized pressure data; The filtered inertial data is normalized to [0, 1], expressed as: ； Where I represents the acceleration data or angular velocity data at any sampling point in the filtered inertial data, Indicates the minimum value of the corresponding acceleration data or angular velocity data I, Indicates the maximum value of the corresponding acceleration data or angular velocity data I, Indicates normalized acceleration data or angular velocity data; The normalized pressure data and inertia data are respectively divided into time windows according to set time, and pressure preprocessing data and inertia preprocessing data are obtained respectively; wherein the inertia preprocessing data includes acceleration preprocessing data or angular velocity preprocessing data.

4. The attention analysis method based on micro-motion detection according to claim 3 is characterized in that: The calculating of the characteristic index value based on the pressure preprocessing data and the inertia preprocessing data includes: Calculate the center of gravity of pressure preprocessing data , expressed as: ； ； in, The horizontal coordinate represents the center of gravity. The ordinate represents the center of gravity. represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and j-th column, M represents the total number of rows, and J represents the total number of columns; Calculate the rate of change of center of gravity , expressed as: ； ； in, represents the rate of change of the horizontal coordinate of the center of gravity position, It represents the rate of change of the vertical coordinate of the center of gravity position, represents the horizontal coordinate of the center of gravity at time t, The ordinate represents the center of gravity at time t; when When it is greater than the set value, a sitting posture change is recorded; the number of sitting posture changes within a period of time is counted to obtain the sitting posture change frequency; Calculate the pressure distribution entropy value H, expressed as: ； in, represents the probability distribution of pressure preprocessing data, , represents the pressure preprocessing data corresponding to the pressure sensor in the i-th row and the j-th column; The pressure preprocessing data of each time window Flatten into a pressure data vector, center the pressure data vectors of all time windows to make their mean zero, and calculate the covariance matrix of the pressure data vector, which is expressed as: ； Where C represents the covariance matrix, E represents the total number of time windows, Indicates The pressure data vector of a time window, T represents transposition; Perform eigenvalue decomposition on the covariance matrix C and obtain the eigenvalue and the corresponding eigenvector ; Select the one with the largest eigenvalue The eigenvectors are used as principal components, and the pressure preprocessing data are projected onto the principal components to obtain dimensional PCA eigenvector of pressure distribution.

5. The attention analysis method based on micro-motion detection according to claim 4 is characterized in that: The calculation of the characteristic index value based on the pressure preprocessing data and the inertia preprocessing data also includes: The mean, variance, kurtosis and skewness of the acceleration preprocessed data or the angular velocity preprocessed data are calculated and expressed as: ； ； ； ； in, Indicates that any axis is Acceleration preprocessing data or angular velocity preprocessing data of a time window, , Represents the x-axis, y-axis and z-axis respectively. represents the mean, E represents the total number of time windows, represents the variance, represents the kurtosis, represents skewness; Perform fast Fourier transform on the acceleration preprocessing data and the angular velocity preprocessing data to obtain corresponding frequency spectra; calculate the energy and frequency center of the acceleration preprocessing data or the angular velocity preprocessing data, which are respectively expressed as: ； ； in, Indicates energy, represents the frequency center, represents the frequency spectrum of acceleration preprocessing data or angular velocity preprocessing data, f represents the fth frequency component in the frequency spectrum, and F represents the total number of frequency components; Calculating the attitude angle and attitude angle change rate through acceleration preprocessing data and angular velocity preprocessing data; wherein the attitude angle includes pitch angle, roll angle and yaw angle; Calculate the amplitude of motion, expressed as: ； in, Indicates the range of motion, Represents the acceleration preprocessing data of the x, y, and z axes respectively.

6. The attention analysis method based on micro-motion detection according to claim 1, characterized in that: The step of inputting the characteristic index value into a pre-trained deep learning model to obtain the attention state classification result of the target object includes: The characteristic index values ​​in each time window are combined into a characteristic vector, and the characteristic vectors in each time window are arranged in chronological order to form a characteristic matrix, which is expressed as: ； Where X represents the feature matrix, Represents the sth characteristic index value in the Eth time window; The feature matrix X is input into the pre-trained deep learning model. In each time window e, the deep learning model receives the feature vector of the current time window. and the hidden state of the previous time window , and output the hidden state of the current time window ; The hidden states of all time windows are input into the deep learning model, and after the hidden states of all time windows are integrated, they are input into the fully connected layer to obtain the corresponding probability values ​​of each attention state, which is expressed as: ； in, represents the probability value of the cth attention state, represents the cth element of the fully connected layer output, and C represents the total number of attention state categories; Based on the corresponding probability value of each attention state, the probability distribution of the attention state is obtained, which is expressed as: ； Where p represents the probability distribution of attention state; The attention state with the largest probability value in the probability distribution p is taken as the attention state classification result.

7. The attention analysis method based on micro-motion detection according to claim 1, characterized in that: The attention trend analysis based on the attention state classification result is performed and the attention change curve is constructed to obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve, including: The attention state classification results in each time window form an attention state sequence, and the duration of each attention state classification result appearing continuously is counted to obtain the duration of each attention state; the number of times each attention state classification result changes within a set time is counted to obtain the conversion frequency of each attention state; The exponential smoothing method is used to smooth the attention state sequence to obtain the attention change curve, which is expressed as: ； in, represents the attention state in the e-th time window, represents the smoothed attention state, represents the smoothing factor; The Pettitt test algorithm is used to detect the mutation points in the attention change curve; Fourier transform was used to analyze whether there was a periodic pattern in the attention change curve.

8. An attention analysis system based on micro-motion detection, characterized in that: include: A data preprocessing module, used to obtain micro-motion data of the target object, and pre-process the micro-motion data to obtain pre-processed data; wherein the micro-motion data includes pressure data and inertia data, and the pre-processed data includes pressure pre-processed data and inertia pre-processed data; A feature calculation module, used for calculating a feature index value based on the pressure preprocessing data and the inertia preprocessing data; A classification module, used to input the characteristic index value into a pre-trained deep learning model to obtain a classification result of the attention state of the target object; The attention analysis module is used to perform attention trend analysis and construct an attention change curve based on the attention state classification result, and obtain the duration and conversion frequency of different attention states and the mutation point of the attention change curve.

9. An attention analysis device based on micro-motion detection, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the attention analysis method based on micro-motion detection as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the attention analysis method based on micro-motion detection described in any one of claims 1 to 7 are implemented.

Images