A millimeter wave-based concentration evaluation method and system
Patent Information
- Application Number
- CN202311780897.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-22
- Publication Date
- 2026-08-21
- Estimated Expiration
- 2043-12-22
AI Technical Summary
但前者会导致不舒适的佩戴体验，视觉摄像头则受限于光线条件和无遮挡等理想情况
[0024] Compared with existing technologies, the present invention has the following advantages: non-contact and low cost; it does not require direct contact with the human body, is highly feasible, has good accuracy, and can achieve convenient evaluation of focus, which helps to provide new technologies and methods for cultivating focus among the educated population.
Smart Images
Figure CN118000730B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of attention detection and recognition technology, and specifically relates to an attention evaluation method and system based on millimeter waves. Background Technology
[0002] Developing focus is crucial in early childhood and primary / secondary education because it directly impacts students' learning behavior and the quality of their learning outcomes. Research shows that children aged 9-12 have poor self-control and difficulty concentrating. Most approach learning with a mindset of simply completing assigned tasks, while their primary focus is on play. This often leads to carelessness, frequent mistakes, low efficiency, and susceptibility to distractions. Over time, children with low focus fail to develop effective learning abilities, negatively affecting their self-confidence, interest in learning, the formation of good study habits, and knowledge accumulation, hindering their healthy development. Therefore, both teachers and parents emphasize cultivating children's focus during class and while completing assignments, both at school and at home. Studies indicate that students with good focus in the classroom learn easily and effectively; students with poor focus are prone to fidgeting, have high error rates in assignments, struggle to improve grades, and not only have low learning efficiency but may also negatively impact the classroom atmosphere and discipline.
[0003] Concentration is a key factor affecting students' learning quality and is also an ability that can be strengthened through training. Cultivating concentration relies on the evaluation of focus levels. Currently, concentration evaluation still mainly depends on human supervision, such as by teachers and parents. Non-human methods for assessing concentration primarily fall into two categories: EEG-based recognition and visual camera recognition. However, the former can lead to an uncomfortable wearing experience, while visual cameras are limited by ideal conditions such as lighting and no obstructions. Summary of the Invention
[0004] In view of this, the primary objective of this invention is to address the shortcomings of existing technologies by proposing a focus evaluation method based on millimeter waves, which can achieve non-contact and sensory perception of human focus.
[0005] The second objective of this invention is to propose a focus evaluation system based on millimeter waves.
[0006] A third objective of this invention is to provide a computer-readable storage medium.
[0007] The first objective of this invention can be achieved through the following technical solution: a focus evaluation method based on millimeter waves, comprising the following steps:
[0008] S1, firstly detects various types of body movement information of the person under different levels of concentration, and defines the relative body movement amplitude to characterize the strength of body movement information;
[0009] S2. Construct a focus evaluation model for individuals under different levels of focus based on relative body movement amplitude;
[0010] S3. Acquire various body movement information of the person being evaluated in real time, and obtain the relative body movement amplitude of the person being evaluated according to step S1. Input the obtained results into the attention evaluation model to obtain the attention prediction results of the person being evaluated.
[0011] Preferably, the detection is performed by measuring the body movement information of a person at different levels of concentration using a non-contact radar sensor.
[0012] Preferably, the relative volumetric amplitude refers to the transformation quantity obtained by taking the first derivative of the ratio of the standard deviations of adjacent distance boxes in the millimeter-wave echo.
[0013] Preferably, the construction of the attention evaluation model includes: compressing the central and outlier features of the relative body movement amplitude through the Huber function dimension to obtain the Huber feature vector representing the state; and then obtaining the attention evaluation model through a nonlinear regressor.
[0014] Preferably, the nonlinear regressor is a support vector regression.
[0015] Preferably, the construction of the attention evaluation model further includes the step of training the nonlinear regression method.
[0016] The second objective of this invention is achieved through the following technical solution: a focus assessment system based on millimeter waves, comprising:
[0017] The personnel movement information collection module is used to collect various types of movement information of personnel at different levels of concentration;
[0018] The relative body movement amplitude generation module is used to calculate the relative body movement amplitude of various body movement information of a person under different levels of concentration;
[0019] The attention evaluation model generation module is used to create attention evaluation models. The relative body movement amplitude is input into the attention evaluation model, which performs a comprehensive attention detection on the subject and obtains thresholds for different attention levels. A nonlinear regressor is used to perform complex threshold divisions and fit attention scores for different attention states. The module continues to increase the number of training objects and state sample points until the performance of the nonlinear regressor reaches the preset accuracy range.
[0020] The focus assessment module is used to input the real-time body movement information of the person being assessed into the trained focus assessment model and to derive the real-time focus score of the person being assessed.
[0021] Preferably, the personnel motion information acquisition module is a radar sensor based on electromagnetic waves, light waves and sound waves. The radar sensor includes a transmitter and a receiver integrated in the same radio frequency front-end module. The receiver is connected to the excitation signal source after passing through a filter, a multiplier and a digital-to-analog converter in sequence. The multiplier is connected to the transmitter after passing through a crystal oscillator and an amplifier in sequence.
[0022] The third objective of the present invention is achieved by the following technical solution: a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method according to the first aspect of the present invention.
[0023] The computer program can be executed by a desktop computer, laptop computer, smartphone, tablet computer, or other terminal device with processor capabilities.
[0024] Compared with existing technologies, the present invention has the following advantages: non-contact and low cost; it does not require direct contact with the human body, is highly feasible, has good accuracy, and can achieve convenient evaluation of focus, which helps to provide new technologies and methods for cultivating focus among the educated population. Attached Figure Description
[0025] Figure 1 This diagram illustrates the application of a radar sensor system in a children's educational setting.
[0026] Figure 2(a) shows the relative body movement amplitude of normal individuals involved in this facility example under high concentration levels (such as resting reading);
[0027] Figure 2(b) shows the relative range of body movements of normal individuals at a moderate level of concentration (e.g., playing with toys at a table) in this facility example.
[0028] Figure 2(c) shows the relative body movement amplitude of a normal person at a low level of concentration (such as back-and-forth swaying) in this embodiment.
[0029] Figure 3 This is a dimensionality-reduced scatter plot of Huber features in eight states of a normal person involved in this embodiment using the t-SNE method.
[0030] Figure 4 For based on Figure 3 The dataset shown is a regressor performance obtained by training with SVR. Detailed Implementation
[0031] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0032] This embodiment provides a millimeter-wave-based attention assessment method, applicable to educational scenarios such as student attention cultivation. This assessment method is implemented through the following millimeter-wave-based attention assessment system. This system specifically includes:
[0033] The personnel movement information collection module is used to collect various types of movement information of personnel at different levels of concentration;
[0034] The relative body movement amplitude generation module is used to calculate the relative body movement amplitude of various body movement information of a person under different levels of concentration;
[0035] The attention evaluation model generation module is used to create attention evaluation models. The relative body movement amplitude is input into the attention evaluation model, which performs a comprehensive attention detection on the subject and obtains thresholds for different attention levels. A nonlinear regressor is used to perform complex threshold divisions and fit attention scores for different attention states. The module continues to increase the number of training objects and state sample points until the performance of the nonlinear regressor reaches the preset accuracy range.
[0036] The focus assessment module is used to input the real-time body movement information of the person being assessed into the trained focus assessment model and to derive the real-time focus score of the person being assessed.
[0037] In specific implementation, such as Figure 1 As shown, the personnel motion information acquisition module is a radar sensor based on electromagnetic waves, light waves, and sound waves. The radar sensor includes a transmitter and a receiver integrated into the same radio frequency front-end module. The receiver uses an orthogonal down-conversion architecture. The receiver is connected to the excitation signal source after passing through a filter, multiplier, and digital-to-analog converter in sequence. The multiplier is then connected to the transmitter after passing through a crystal oscillator and amplifier in sequence. The transmitter emits electromagnetic waves towards the person being assessed, and these waves are reflected by the person's body and received by the radar receiver.
[0038] Based on the above principles, the present invention provides a focus evaluation method based on millimeter waves, comprising the following steps:
[0039] S1. First, the radar sensor in the aforementioned personnel motion information acquisition module measures and detects various types of personnel motion information under different levels of concentration. The standard deviation of the radar echo within each distance box is measured by the radar sensor, and the relative motion amplitude is defined to characterize the strength of the motion information.
[0040] In this step, the method for measuring the standard deviation of radar echoes is as follows:
[0041] First, a Fourier transform is performed on the demodulated signal to calculate the distance bins according to each frequency bin. The distance d is related to the frequency f.m The correspondence is as follows:
[0042]
[0043] In the formula: c is the speed of light, in m / s, and K is the frequency modulation slope, in Hz / s.
[0044] In practice, the distance-to-bin spacing is also related to the sampling rate. Under continuous sampling conditions, the optimal distance resolution is only related to the sweep bandwidth, i.e., the minimum distance-to-bin spacing is:
[0045]
[0046] In the formula: BW is the sweep bandwidth, in Hz.
[0047] Within each distance box interval, the standard deviation of the signal falling within it can be calculated, i.e.:
[0048]
[0049] In the formula: Let x be the average amplitude of n signals collected within the distance box. i Let represent the amplitude of the i-th signal in the statistics, and n be the total number of signal amplitudes calculated within the distance box.
[0050] The strength of a human presence within each bin can be characterized by its standard deviation (Std). When the human body moves relative to each bin, the ratio of the standard deviations between adjacent bins will differ. To avoid directional confusion, one way to define the ratio of the standard deviations of two adjacent bins is as follows:
[0051]
[0052] The defined standard deviation ratio represents the magnitude of an object's jump amplitude when jumping between distance boxes. When there is no relative body movement, this ratio is close to 2; when there is relative body movement, this ratio is positively correlated with the jump amplitude. Based on the above definition, the relative body movement amplitude of a person will exhibit different differences under different levels of concentration; the lower the concentration, the greater the relative body movement amplitude.
[0053] As shown in Figure 2(a), at a high level of concentration, the relative body movement amplitude values of the individuals are all less than 1.
[0054] As shown in Figure 2(b), at a moderate level of concentration, the relative body movement amplitude of the person being represented will exhibit an outlier value of 20.
[0055] As shown in Figure 2(c), at low levels of focus, the relative body movement amplitude of the person being represented exhibits an outlier value as high as 75.
[0056] To reduce noise interference and to represent the absence of body motion amplitude with 0, we can take the first derivative of this ratio, i.e., find the difference. This leads to a definition of relative body motion amplitude as follows:
[0057] Δx=(ratio)′·α
[0058] In the definition of relative body movement amplitude, α is a scaling factor, which defaults to 1. It can be used to scale the amplitude range to better distinguish differences in attention score regression across different states. It's important to note that this scaling factor does not change the description of relative body movement amplitude, as it uses the same scaling factor.
[0059] S2. Construct a focus evaluation model for individuals under different levels of focus based on relative body movement amplitude;
[0060] By collecting data from normal individuals in various states, it can be observed that in a state of inattention, individuals exhibit increased fidgeting and behaviors such as playing around at a table. The greater the amplitude of these movements, the more inattentive the individual is defined as. It can be found that as the degree of inattention increases, the relative amplitude of body movements exhibits two characteristics: firstly, the amplitude of concentration values increases; and secondly, there is a greater probability of large outliers appearing.
[0061] Since the relative body motion amplitude data is a time-domain sampling sequence, although a direct score evaluation can be performed by setting a certain jump threshold, this evaluation is too coarse and easily affected by outliers. Furthermore, when retaining data for a certain duration for score regression analysis, the original data dimensionality is too high. Therefore, it is necessary to construct features after data compression. In this embodiment, the data compression method is the Huber function. In machine learning, the Huber function is used to consider the influence of concentrated features while retaining outliers. Therefore, the Huber function is chosen to perform dimensionality reduction while accumulating feature values. Its definition is as follows:
[0062]
[0063] The Huber function is a continuous, first-order differentiable function. Here, δ is the threshold value in the Huber function, and it can be seen that in the low-value part x... i The feature will be reduced between 0 and 1, which means the suppression of the case without body motion amplitude, while the feature will be amplified between 1 and δ, which means the emphasis on the first characteristic, that is, the concentration value after the amplitude is increased is amplified, and in the case greater than δ, it means the preservation of the second characteristic, that is, the influence of outliers is also effective.
[0064] Using δ as the partition, the eigenvalues within each of the two partitions are summed to obtain their respective eigenvector representations. Thus, after passing through the Huber function, the high-dimensional data sequence of a certain duration is compressed into two-dimensional eigenvectors. If k distance bin ratios are selected, then 2^k eigenvectors will be constructed. In general, k is usually less than 5, therefore the Huber function achieves feature compression of high-dimensional data, and the compressed features are simply called Huber features.
[0065] By collecting data on a person's state for a certain period of time, and further compressing the calculated relative body motion amplitude using the Huber function dimension, a Huber feature vector representing that state is obtained. This feature vector takes the form:
[0066]
[0067] Where Δx i For relative body motion amplitude less than δ, Δx j The relative volumetric motion amplitude is greater than δ, and N is the total number of points in the selected sequence, which is related to the sampling rate f. s It is related to the selected time length T, that is, N = f s T, k is the number of adjacent distance bin ratios selected. For example... Figure 3 As shown, t-SNE dimensionality reduction is performed on the Huber features obtained under different states, where k is chosen to be 3. It can be observed that under low, medium, and high attention levels, data points tend to cluster at different locations on the planar graph. Two rough threshold lines are drawn in the graph to represent this difference. Therefore, a nonlinear regressor is used to achieve more complex threshold line divisions, thereby fitting attention scores for different attention states.
[0068] Among numerous nonlinear regressors, Support Vector Regression (SVR) is chosen to evaluate the regression scores of data points under various states. The mathematical definition of SVR is "minimizing the distance to the sample point farthest from the hyperplane," which involves solving the following optimization problem:
[0069]
[0070] In the formula, w is the weight vector of the regression model, b is the bias term of the regression model, and y i Let x be the label value of the i-th data item. i Let be the feature vector of the i-th data point, ε be the tolerance range of deviation, and ε be a constant.
[0071] In the context of a linear problem, its optimal solution is a linear boundary condition f(x) = wx + b. By introducing a kernel function, the nonlinear boundary condition of the nonlinear problem can be further obtained, i.e., the optimal solution becomes f(x) = wx + b. T The relationship between the kernel function κ and the resulting low-dimensional to high-dimensional mapping φ is:
[0072] κ(x i ,x j )=φ(x i ) T φ(x j )
[0073] This embodiment uses a Gaussian kernel function, also known as a radial basis function, which maps each sample point to an infinite-dimensional feature space, making it suitable for situations with a small number of sample points. Its definition is as follows:
[0074]
[0075] In the above formula, σ is the defined Gaussian kernel bandwidth.
[0076] S3. Acquire various body movement information of the person being evaluated in real time, and obtain the relative body movement amplitude of the person being evaluated according to step S1. Input the obtained results into the attention evaluation model to obtain the attention prediction results of the person being evaluated.
[0077] Inputting this feature vector into a pre-trained support vector regressor will yield a score evaluation for that state. That is:
[0078] f(X) = w T φ(X)+b
[0079] In the formula, w T Let be the weight vector of the regression model trained in S2, b be the bias term of the regression model trained in S2, φ be the mapping defined by the kernel function, and X be the feature vector of the new input data. Figure 3 The dataset shown, Figure 4 The dataset was trained and tested. The results show that, defining 75 points as the high focus threshold and 50 points as the low focus threshold, the system can provide training scores for each state based on prior scores. For new data in the untrained typing state, the average score is 78.70, falling within the defined high focus range, thus demonstrating that the support vector regressor based on the training data has an appropriate scoring ability for new data. By training with more sample points from more objects and states, the performance of the regressor can be further improved.
[0080] Furthermore, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the millimeter-wave-based attention evaluation method according to Embodiment 1. The computer program described in this embodiment can be executed by a desktop computer, laptop computer, smartphone, tablet computer, or other terminal device with processor functionality.
[0081] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.
Claims
1. A focus assessment method based on millimeter waves, characterized in that, Includes the following steps: S1, firstly detects various types of body movement information of the person under different levels of concentration, and defines the relative body movement amplitude to characterize the strength of body movement information; S2. Construct a focus evaluation model for personnel under different levels of focus based on the relative body movement amplitude. The construction of the focus evaluation model includes: compressing the central and outlier features of the relative body movement amplitude through the Huber function dimension to obtain the Huber feature vector representing the state. Then, a focus evaluation model is obtained through a nonlinear regressor. S3. Acquire various body movement information of the person being evaluated in real time, and obtain the relative body movement amplitude of the person being evaluated according to step S1. Input the obtained results into the attention evaluation model to obtain the attention prediction result of the person being evaluated. The relative body movement amplitude refers to the transformation quantity obtained by taking the first derivative of the ratio of the standard deviations of adjacent distance boxes in the millimeter wave echo.
2. The focus evaluation method based on millimeter waves according to claim 1, characterized in that, The detection method uses a non-contact radar sensor to measure the body movement information of a person at different levels of concentration.
3. A focus evaluation method based on millimeter waves according to claim 1 or 2, characterized in that, The nonlinear regressor is a support vector regression.
4. A focus evaluation method based on millimeter waves according to claim 1 or 2, characterized in that, The construction of the attention evaluation model also includes the step of training the nonlinear regressor.
5. A focus assessment device based on millimeter waves, comprising: The personnel movement information collection module is used to collect various types of movement information of personnel at different levels of concentration; The relative body motion amplitude generation module is used to calculate the relative body motion amplitude of various body motion information of a person under different levels of concentration. The relative body motion amplitude refers to the transformation quantity obtained by taking the first derivative of the ratio of the standard deviations of adjacent distance boxes in the millimeter wave echo. The focus assessment model generation module is used to create focus assessment models. The relative body movement amplitude is input into the focus assessment model, and the focus assessment model performs a comprehensive focus test on the subject to obtain thresholds for different focus levels. A nonlinear regressor is used to perform complex threshold division and fit focus scores for different focus levels. And by increasing the number of training objects and state sample points, the nonlinear regressor performance is increased until it reaches the preset accuracy range. The construction of the attention evaluation model includes: compressing the central and outlier features of the relative body movement amplitude through the Huber function dimension to obtain the Huber feature vector representing the state; Then, a focus evaluation model is obtained through a nonlinear regressor. The focus assessment module is used to input the real-time body movement information of the person being assessed into the trained focus assessment model and to derive the real-time focus score of the person being assessed.
6. The attention evaluation device based on millimeter waves according to claim 5, characterized in that, The personnel motion information acquisition module is a radar sensor based on electromagnetic waves, light waves and sound waves. The radar sensor includes a transmitter and a receiver integrated in the same radio frequency front-end module. The receiver is connected to the excitation signal source after passing through a filter, a multiplier and a digital-to-analog converter in sequence. The multiplier is connected to the transmitter after passing through a crystal oscillator and an amplifier in sequence.
7. A computer-readable storage medium having a computer program stored thereon, the computer program implementing the method of any one of claims 1-4 when executed by a processor.
8. A computer-readable storage medium according to claim 7, characterized in that, The computer program is executed by a desktop computer, laptop computer, smartphone, tablet computer, or other terminal device with processor functionality.