Psychological state assessment method and device

Abstract

The invention discloses a psychological state assessment method and device. The method comprises the steps that firstly, based on a sensor group and a camera module, electrocardio information, face information, gait information and plantar pressure information of a target are obtained; then constructing an input feature vector based on the electrocardio information, the face information, the gait information and the plantar pressure information; and finally, inputting the input feature vector into a preset model to obtain an evaluation result. According to the embodiment of the invention, the multi-dimensional feature data can be processed, so that the psychological state is evaluated by using the multi-dimensional feature data to obtain the evaluation score, and the evaluation accuracy can be improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a method and apparatus for assessing psychological state. Background Technology

[0002] In terms of psychological state assessment, traditional methods generally rely on psychologists or neural networks such as LSTM for assessment. However, existing technologies such as LSTM use limited data and have low assessment accuracy.

[0003] Therefore, how to use multidimensional feature data to more accurately assess psychological state is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problem of the accuracy of psychological state assessment caused by the use of single data in the prior art.

[0005] To achieve the above-mentioned technical objectives, in one aspect, the present invention provides a method for assessing psychological states, the method comprising: The sensor array and camera module are used to acquire the target's electrocardiogram information, facial information, gait information and plantar pressure information; An input feature vector is constructed based on the electrocardiogram information, facial information, gait information, and plantar pressure information. The input feature vector is input into a preset model to obtain the evaluation result.

[0006] Furthermore, the electrocardiogram information includes pre-exercise electrocardiogram information, exercise electrocardiogram information, and post-exercise electrocardiogram information; the facial information includes pre-exercise facial information, exercise facial information, and post-exercise facial information; the gait information includes pre-exercise gait information, exercise gait information, and post-exercise gait information; and the plantar pressure information includes pre-exercise plantar pressure information, exercise plantar pressure information, and post-exercise plantar pressure information.

[0007] Furthermore, the construction of the input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information specifically includes: Based on the electrocardiogram information, facial information, gait information, and plantar pressure information, construct the difference characteristics and relative change characteristics between each pair of the pre-exercise stage, the exercise stage, and the post-exercise stage; The input feature vector is obtained by concatenating the difference features, relative change features, and original stage features. The original stage features specifically refer to the features of the pre-motion stage, the mid-motion stage, and the post-motion stage.

[0008] Furthermore, the preset model is specifically an extreme gradient boosting regression model, and the objective function of the preset model during training is as follows:

[0009] In the formula, The overall objective function value of the preset model during training is composed of the loss term and the regularization term. The total number of training samples. Let the mean squared error loss function be . This represents the true psychological state score of the i-th training sample. Let be the predicted value given by the model for the i-th training sample after the t-th iteration. This represents the current iteration round number or the current accumulated number of regression trees. This is the complexity penalty term for the k-th regression tree.

[0010] Furthermore, the aforementioned Specifically, it is determined using the following formula:

[0011] In the formula, Let be the input feature vector of the i-th training sample.

[0012] Furthermore, the method also includes, in each round of training, the loss function of the preset model as follows:

[0013] In the formula, This is the approximate objective function value obtained by performing a second-order Taylor expansion on the original loss function at the t-th iteration. Let be the first-order gradient of the i-th training sample under the current prediction, i.e., the partial derivative of the loss function with respect to the predicted value. The regression tree fitted for this round The output value of the i-th training sample, Let be the second gradient of the i-th training sample, i.e., the second partial derivative of the loss function with respect to the predicted value. Let be the regularization term for the t-th regression tree, used to limit the complexity of the tree.

[0014] Furthermore, the method also includes determining the average information gain of each feature at all nodes across all trees. This average information gain is used to quantify the contribution of the corresponding feature in the preset model, and is expressed by the following formula:

[0015] In the formula, For average information gain, The total number of times a feature is split is a characteristic. The loss function value before the feature is split. The loss function value is the value after the feature is split.

[0016] On the other hand, the present invention also provides a psychological state assessment device, the device comprising: The acquisition module, including a sensor group and a camera module, is used to acquire the target's electrocardiogram information, facial information, gait information, and plantar pressure information; The construction module is used to construct an input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information; The evaluation module is used to input the input feature vector into a preset model to obtain the evaluation result.

[0017] This invention provides a method and apparatus for assessing psychological state. Compared with existing technologies, this method first acquires the target's electrocardiogram (ECG), facial information, gait information, and plantar pressure information based on a sensor array and a camera module. Then, it constructs an input feature vector based on the ECG, facial, gait, and plantar pressure information. Finally, it inputs the feature vector into a preset model to obtain the assessment result. This method can process multi-dimensional feature data, thereby using it to assess psychological state and obtain an assessment score, thus improving the accuracy of the assessment. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 The diagram shown is a flowchart illustrating the psychological state assessment method provided in the embodiments of this specification. Figure 2 The diagram shown is a structural schematic of the psychological state assessment device provided in the embodiments of this specification. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] like Figure 1 The diagram illustrates a flowchart of the psychological state assessment method provided in the embodiments of this specification. Although this specification provides the method operation steps or device structure shown in the following embodiments or figures, based on convention or without creative effort, the method or device may include more or fewer operation steps or module units after partial merging. In steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device are not limited to the execution order or module structure shown in the embodiments or figures of this specification. When the method or module structure is applied in actual devices, servers, or terminal products, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiments or figures (e.g., in a parallel processor or multi-threaded processing environment, or even in a distributed processing or server cluster implementation environment).

[0022] The psychological state assessment methods provided in the embodiments of this specification can be applied to terminal devices such as client and server devices, such as... Figure 1 As shown, the method specifically includes the following steps: Step S101: Obtain the target's electrocardiogram information, facial information, gait information and plantar pressure information based on the sensor group and camera module.

[0023] Specifically, the sensors include accelerometers and velocity sensors. Combined with video acquired by the camera module, the sensor array can obtain gait information of the target user, including stride length, stride width, average gait speed, gait cycle variation coefficient, left-right support ratio, and facial information. It focuses on extracting eye-related events (e.g., blink frequency, blink amplitude), mouth activities (e.g., opening the mouth, smiling, lip tremors), and head movement parameters (e.g., left-right or up-down swing angles, trembling). These features reflect changes in facial behavior under different psychological and emotional states; for example, anxiety may manifest as frequent blinking or mouth opening, and increased head movement. The sensor array also includes a plantar pressure sensor array, with 16 independent pressure detection units per foot, covering the forefoot, arch, and heel areas. The sensors have an accuracy of 0.1N and a refresh rate of 100Hz, accurately capturing the pressure distribution trajectory and trends within the gait cycle. This module transmits data to the computing unit in real-time via Bluetooth communication and supports synchronous timestamp calibration to ensure consistency with other modal data in the temporal dimension. Communication between the computing unit and the camera is via USB, communication between the computing unit and the physiological signal acquisition module is via Bluetooth, and communication between the computing unit and the plantar pressure acquisition module is via Bluetooth. The plantar pressure information specifically includes the offset of the plantar center trajectory and the maximum pressure value.

[0024] In heart rate variability (HRV) analysis, the ECG signal was first filtered and R-waves were detected to extract all normal RR intervals. The following indices were then calculated: the standard deviation of all normal RR intervals (SDNN); the root mean square (RMSSD) of the difference between adjacent RR intervals; the percentage (PNN50) of adjacent RR intervals with a difference exceeding 50 ms; the standard deviation (SDANN) of the average RR interval every 5 minutes; high-frequency power (HF) (reflecting parasympathetic activity), low-frequency power (LF) (reflecting combined sympathetic and parasympathetic regulation), very low-frequency power (VLF), and the LF / HF ratio; as well as sample entropy and fractal exponent, reflecting signal complexity. Additionally, Poincaré plot parameters SD1 and SD2 were calculated, representing short-term and long-term heart rate variability, respectively. All HRV features were calculated using data from three phases: initial rest (the first 2 minutes of the resting phase), late exercise (the last 2 minutes of the exercise phase), and final rest (the last 2 minutes of the recovery rest phase) to create multi-time-segment comparative features.

[0025] The electrocardiogram (ECG) information includes pre-exercise ECG information, exercise ECG information, and post-exercise ECG information; the facial information includes pre-exercise facial information, exercise facial information, and post-exercise facial information; the gait information includes pre-exercise gait information, exercise gait information, and post-exercise gait information; and the plantar pressure information includes pre-exercise plantar pressure information, exercise plantar pressure information, and post-exercise plantar pressure information.

[0026] Specifically, each piece of information includes information from three phases: the initial rest phase (before exercise, the first 2 minutes of the rest phase), the exercise phase (the last 2 minutes of the exercise phase), and the end rest phase (after exercise, the last 2 minutes of the recovery rest phase), in order to form a comparative feature across multiple time periods.

[0027] Step S102: Construct an input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information.

[0028] The construction of the input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information specifically includes: Based on the electrocardiogram information, facial information, gait information, and plantar pressure information, construct the difference characteristics and relative change characteristics between each pair of the pre-exercise stage, the exercise stage, and the post-exercise stage; The input feature vector is obtained by concatenating the difference features, relative change features, and original stage features. The original stage features specifically refer to the features of the pre-motion stage, the mid-motion stage, and the post-motion stage.

[0029] Specifically, the input features are a combination vector consisting of three modalities, including: Heart rate variability (HRV) features include 12 items: SDNN, RMSSD, PNN50, SDANN, HF, LF, VLF, LF / HF, sample entropy, fractal dimension, SD1, and SD2. ; Video facial behavioral characteristics include six items: blink count per unit time, duration of eye closure, mouth opening frequency, mouth amplitude, head pitch rate, and horizontal rotation frequency. ; Gait and plantar pressure characteristics include seven items: stride length, stride width, average gait speed, gait cycle variation coefficient, left-right support ratio, plantar center trajectory deviation, and maximum pressure value. .

[0030] Statistical values ​​(such as mean, standard deviation, etc.) of the above features are extracted for each stage (pre-resting stage S1, post-exercise stage S2, post-resting stage S3), and stage difference features are further constructed: Difference characteristics: ; Relative rate of change characteristics: .

[0031] Ultimately, the model's input vector is a concatenation of all original stage features and stage difference features:

[0032] The total dimension of this vector is 9×(12+6+7)=2259×(12+6+7)=225. By introducing this cross-stage dynamic change index, the system can effectively capture the impact of exercise-induced changes in physical and mental state on psychological reactions, thereby constructing a more discriminative input vector in the time dimension, significantly enhancing the representation ability and discriminativeness of features and the sensitivity of the model to state transitions.

[0033] Step S103: Input the input feature vector into the preset model to obtain the evaluation result.

[0034] The preset model is specifically an extreme gradient boosting regression model, specifically an extreme gradient boosting regression model (XGBoost). This algorithm is an efficient implementation of the Boosting framework, with advantages such as fast computation speed, strong model generalization ability, insensitivity to feature scale, and built-in regularization mechanism. It is particularly suitable for nonlinear regression tasks of multimodal high-dimensional data.

[0035] The core of XGBoost is to process the feature vector of each sample. It is directly used as the input to each regression tree. That is, the existing input vector X (from all the trees in the training set) is used as the input. The matrix is ​​fed into XGBoost row by row: the model divides samples into different leaves by finding binary splitting conditions on features, and each leaf has a corresponding fitted value (leaf weight). Formally, the prediction is... ,in It is the regression tree function space. Let be the input feature vector of the i-th training sample. The loss function is the mean squared error (MSE): In addition, there is a tree complexity penalty term. (Common format, Penalty based on the number of leaves. L2 penalty for leaf weights, The number of leaves, (This is the leaf weight vector). The feature vector X of each sample serves as the basis for tree partitioning; during tree construction, the split point of each feature is examined and the loss reduction caused by the split is calculated, thereby growing the tree structure and obtaining the evaluation result.

[0036] The overall model structure is based on the additive model and the forward stepwise optimization principle. The objective function of the preset model is shown in the following formula:

[0037] In the formula, The overall objective function value of the preset model during training is composed of the loss term and the regularization term. The total number of training samples. Let the mean squared error loss function be . This represents the true psychological state score of the i-th training sample. Let be the predicted value given by the model for the i-th training sample after the t-th iteration. This represents the current iteration round number or the current accumulated number of regression trees. This is the complexity penalty term for the k-th regression tree.

[0038] In this embodiment of the application, the method further includes pre-training a preset model, and in each round of training, performing a second-order approximation of the loss function through Taylor expansion, and utilizing the first-order derivative. and second derivative The loss function can be approximated as:

[0039] In the formula, This is the approximate objective function value obtained by performing a second-order Taylor expansion on the original loss function at the t-th iteration. Let be the first-order gradient of the i-th training sample under the current prediction, i.e., the partial derivative of the loss function with respect to the predicted value. The regression tree fitted for this round The output value of the i-th training sample, Let be the second gradient of the i-th training sample, i.e., the second partial derivative of the loss function with respect to the predicted value. Let be the regularization term for the t-th regression tree, used to limit the complexity of the tree.

[0040] This constructs the optimal regression tree structure for each round, resulting in a more refined fit to the residuals. To further optimize model performance, this invention employs the following parameter configurations: Basic learner type: gbtree (CART type regression tree); Number of trees (n_estimators): 200; Maximum tree depth (max_depth): 6; Learning rate (eta): 0.05; Minimum splitting loss (gamma): 0.1; L2 regularization term (lambda): 1; Subsample ratio: 0.8; Column sampling ratio (colsample_bytree): 0.8; Maximum number of leaf nodes: 64.

[0041] During model training, a 5-fold cross-validation strategy is adopted, in which 80% of the samples are used for training and 20% for validation on the training set. The mean squared error of the validation set is monitored at each fold, and the model performance is finally calculated by averaging to prevent overfitting and improve model stability.

[0042] In this embodiment, to improve the model's adaptability to the importance of multimodal features, the present invention introduces a weighted mechanism based on the dynamic feature "gain" ranking during the training process: this mechanism is tightly integrated into each iteration of the aforementioned XGBoost algorithm. When constructing each regression tree, the algorithm needs to select the optimal feature for splitting at each node to minimize the loss function. The system records the average information gain (Gain) of each feature at all splitting nodes across all trees—that is, the average reduction in the loss function when the feature is used for splitting—as shown in the following formula:

[0043] In the formula, For average information gain, The total number of times a feature is split is a characteristic. The loss function value before the feature is split. The loss function value is the value after the feature is split.

[0044] Furthermore, average information gain is used to quantify the contribution of each feature to the model. Specific uses of average information gain include: ① Feature importance ranking: Features are ranked by calculating their average information gain. Features with high gain indicate that they are more effective at reducing prediction errors and contributing more to the assessment of psychological states in the model. (That is, the importance is ranked based on the average gain accumulated over historical iterations.) ② Feature selection and optimization: When the number of features is high or redundant, low-contribution features are removed based on average information gain, or their sampling probability is reduced during training, improving model training efficiency and reducing the risk of overfitting. ③ Dynamic adjustment of feature weights: In incremental training or multi-stage feature fusion scenarios, the weights of features in subsequent tree construction or stage feature fusion are dynamically adjusted based on average information gain, making the model focus more on effective information.

[0045] During training, the system ranks the importance of each feature based on its accumulated average gain over historical iterations. Features with consistently poor performance and low gain (i.e., "low-gain features") are either downweighted (e.g., their sampling probability is reduced) or directly eliminated (e.g., features with gain below a set threshold are ignored). This mechanism works in conjunction with XGBoost's regularization framework to partially simulate the effect of an "attention mechanism" without introducing deep neural networks. This allows the model to adaptively focus on features with greater information content, thereby significantly improving robustness and interpretability in heterogeneous, high-dimensional, and imbalanced feature scenarios.

[0046] This proposed solution overcomes the limitations of traditional psychological state assessments that rely on single data sources, subjective questionnaires, or static behavioral observations, achieving dynamic, objective, and accurate assessment of individual psychological states. In experimental validation, the system achieved an accuracy rate exceeding 90% in identifying anxious and non-anxious groups, improving upon traditional single-modal methods based on heart rate or facial expressions by more than 15%, significantly enhancing the sensitivity and specificity of the assessment. It can be applied to scenarios such as pre-competition psychological adaptability assessment for athletes, psychological state monitoring for medical rehabilitation patients, and stress load screening for occupational groups, and possesses good data interpretability, real-time response capabilities, and significant potential for widespread application.

[0047] Based on the above-described psychological state assessment method, one or more embodiments of this specification also provide a platform or terminal for psychological state assessment. This platform or terminal may include devices, software, modules, plug-ins, servers, clients, etc., using the methods described in the embodiments of this specification, combined with necessary hardware implementation devices. Based on the same innovative concept, the systems in one or more embodiments provided in this specification are as described in the following embodiments. Since the implementation schemes and methods for solving the system problem are similar, the specific system implementation in the embodiments of this specification can refer to the implementation of the aforementioned methods. Repeated descriptions will not be repeated. The terms "unit" or "module" used below can refer to a combination of software and / or hardware that achieves a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, hardware implementation, and a combination of software and hardware, are also possible and contemplated.

[0048] Specifically, Figure 2 This is a schematic diagram of the module structure of one embodiment of the psychological state assessment device provided in this specification, as shown below. Figure 2 As shown, the psychological state assessment device provided in this instruction manual includes: The acquisition module 201 includes a sensor group and a camera module, used to acquire the target's electrocardiogram information, facial information, gait information and plantar pressure information; Construction module 202 is used to construct an input feature vector based on the electrocardiogram information, facial information, gait information and plantar pressure information; The evaluation module 203 is used to input the input feature vector into a preset model to obtain the evaluation result.

[0049] It should be noted that the system described above may include other implementation methods based on the description of the corresponding method embodiments. The specific implementation methods can be referred to the description of the corresponding method embodiments above, and will not be elaborated here.

[0050] This application also provides an electronic device, including: processor; Memory used to store the processor's executable instructions; The processor is configured to perform the methods provided in the embodiments described above.

[0051] The electronic device provided in this application stores executable instructions for a processor in a memory. When the processor executes these instructions, it first acquires electrocardiogram (ECG), facial, gait, and plantar pressure information of a target based on a sensor array and a camera module. Then, it constructs an input feature vector based on the ECG, facial, gait, and plantar pressure information. Finally, it inputs the input feature vector into a preset model to obtain an evaluation result. This device can process multi-dimensional feature data to evaluate psychological states and obtain evaluation scores, thereby improving the accuracy of the evaluation.

[0052] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0053] The methods or apparatus described in the embodiments provided in this specification can implement business logic through a computer program and record it on a storage medium. The storage medium can be read and executed by a computer to achieve the effects of the solutions described in the embodiments of this specification, such as: The sensor array and camera module are used to acquire the target's electrocardiogram information, facial information, gait information and plantar pressure information; An input feature vector is constructed based on the electrocardiogram information, facial information, gait information, and plantar pressure information. The input feature vector is input into a preset model to obtain the evaluation result.

[0054] The storage medium can include physical devices for storing information, typically digitizing the information and then storing it using electrical, magnetic, or optical methods. The storage medium can include: devices that store information using electrical energy, such as various types of memory, like RAM and ROM; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memory, bubble memory, and USB flash drives; and devices that store information using optical methods, such as CDs or DVDs. Of course, there are other readable storage media, such as quantum memories and graphene memories.

[0055] The embodiments in this specification are not limited to conforming to industry communication standards, standard computer resource data update and data storage rules, or the situations described in one or more embodiments of this specification. Slightly modified implementations based on certain industry standards or custom methods or embodiments can also achieve the same, equivalent, or similar, or predictable, implementation effects as described above. Embodiments that utilize these modified or modified methods for data acquisition, storage, judgment, and processing still fall within the scope of optional implementations of the embodiments in this specification.

[0056] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, ASICs, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0057] The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or plug-ins may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0058] These computer program instructions can also be loaded onto a computer or other programmable resource data updating device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, system embodiments are basically similar to method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. In the description of this specification, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this specification. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0060] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for assessing psychological state, characterized in that, The method includes: The sensor array and camera module are used to acquire the target's electrocardiogram information, facial information, gait information and plantar pressure information; An input feature vector is constructed based on the electrocardiogram information, facial information, gait information, and plantar pressure information. The input feature vector is input into a preset model to obtain the evaluation result.

2. The psychological state assessment method as described in claim 1, characterized in that, The electrocardiogram (ECG) information includes pre-exercise ECG information, exercise ECG information, and post-exercise ECG information; the facial information includes pre-exercise facial information, exercise facial information, and post-exercise facial information; the gait information includes pre-exercise gait information, exercise gait information, and post-exercise gait information; and the plantar pressure information includes pre-exercise plantar pressure information, exercise plantar pressure information, and post-exercise plantar pressure information.

3. The psychological state assessment method as described in claim 2, characterized in that, The construction of the input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information specifically includes: Based on the electrocardiogram information, facial information, gait information, and plantar pressure information, construct the difference characteristics and relative change characteristics between each pair of the pre-exercise stage, the exercise stage, and the post-exercise stage; The input feature vector is obtained by concatenating the difference features, relative change features, and original stage features. The original stage features specifically refer to the features of the pre-motion stage, the mid-motion stage, and the post-motion stage.

4. The psychological state assessment method as described in claim 1, characterized in that, The preset model is specifically an extreme gradient boosting regression model, and the objective function of the preset model during training is shown in the following formula: In the formula, The overall objective function value of the preset model during training is composed of the loss term and the regularization term. The total number of training samples. Let the mean squared error loss function be . This represents the true psychological state score of the i-th training sample. Let be the predicted value given by the model for the i-th training sample after the t-th iteration. This represents the current iteration round number or the current accumulated number of regression trees. This is the complexity penalty term for the k-th regression tree.

5. The psychological state assessment method as described in claim 4, characterized in that, The Specifically, it is determined using the following formula: In the formula, Let be the input feature vector of the i-th training sample.

6. The psychological state assessment method as described in claim 4, characterized in that, The method further includes, in each round of training, the loss function of the preset model as follows: In the formula, This is the approximate objective function value obtained by performing a second-order Taylor expansion on the original loss function at the t-th iteration. Let be the first-order gradient of the i-th training sample under the current prediction, i.e., the partial derivative of the loss function with respect to the predicted value. The regression tree fitted for this round The output value of the i-th training sample, Let be the second gradient of the i-th training sample, i.e., the second partial derivative of the loss function with respect to the predicted value. Let be the regularization term for the t-th regression tree, used to limit the complexity of the tree.

7. The psychological state assessment method as described in claim 6, characterized in that, The method further includes determining the average information gain of each feature at all nodes of all trees. The average information gain is used to quantify the contribution of the corresponding feature in the preset model. The average information gain is shown in the following formula: In the formula, For average information gain, The total number of times a feature is split is a characteristic. The loss function value before the feature is split. The loss function value is the value after the feature is split.

8. A psychological state assessment device, characterized in that, The device includes: The acquisition module, including a sensor group and a camera module, is used to acquire the target's electrocardiogram information, facial information, gait information, and plantar pressure information; The construction module is used to construct an input feature vector based on the electrocardiogram information, facial information, gait information, and plantar pressure information; The evaluation module is used to input the input feature vector into a preset model to obtain the evaluation result.

Images