Air rifle athlete training emotion monitoring method based on phase entanglement cooperation

Abstract

The invention provides an air rifle athlete training emotion monitoring method based on phase entanglement collaboration, which is used for realizing deep fusion of context perception on the basis of deep spatio-temporal features extracted from electroencephalogram signals and facial micro-expressions by a phase entanglement collaboration network and a micro-expression recognition model. Therefore, the complex emotional state of the air rifle athlete can be synchronously and accurately recognized in a high-pressure and fast-changing training environment, and then high-precision data support is provided for monitoring and intervention of the emotional state of the air rifle athlete.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, specifically to a method for monitoring the training emotions of air rifle athletes based on phase entanglement collaboration. Background Technology

[0002] An athlete's training state and emotions under high-pressure competition often directly affect their decision-making, stability and final performance. This is especially true in sports such as the 10-meter air rifle, which require a high degree of psychological stability. Traditional training methods often focus on technical movements and physical fitness, lacking real-time and objective monitoring of the air rifle athlete's internal emotional state. This makes it difficult to provide accurate psychological intervention and guidance at critical moments.

[0003] In recent years, emotion computing technology based on physiological and behavioral signals has provided a new approach for assessing the training status of air rifle athletes.

[0004] However, the inventors of this application have found that the existing methods still have limitations, as they rely on single-modal data and are difficult to fully capture the subtle and rapidly changing emotional states of air rifle athletes under high pressure. Summary of the Invention

[0005] This application provides a method for monitoring the training emotions of air rifle athletes based on phase entanglement collaboration. It uses deep spatiotemporal features extracted from EEG signals and facial micro-expressions by a phase entanglement collaboration network and a micro-expression recognition model to achieve deep fusion of context awareness. This enables synchronous and accurate identification of the complex emotional states of air rifle athletes in high-pressure and rapidly changing training environments, thereby providing high-precision data support for the monitoring and intervention of the emotional states of air rifle athletes.

[0006] Firstly, this application provides a method for monitoring training emotions of air rifle athletes based on phase entanglement collaboration, the method comprising: Acquire EEG signals and facial videos of current air rifle athletes during training; After temporally aligning EEG signals and facial videos, the brain activity and emotional results corresponding to the EEG signals are extracted through a phase entanglement collaborative network, and the emotional facial representations corresponding to the facial videos are extracted through a micro-expression recognition model. Based on the multimodal features obtained by fusing brain activity emotional outcomes and emotional facial representations, a phased dynamic assessment of emotional state is conducted. Based on the phased dynamic assessment results and the pre-set emotional state classification table, the current emotional state of the air rifle athlete is determined.

[0007] Secondly, this application provides a phase entanglement-based collaborative training emotion monitoring device for air rifle athletes, the device comprising: The acquisition unit is used to acquire the electroencephalogram (EEG) signals and facial videos of the current air rifle athlete during training. The extraction unit is used to extract the brain activity and emotional results corresponding to the EEG signals through a phase entanglement collaborative network after performing temporal alignment operations on EEG signals and facial videos, and to extract the emotional facial representations corresponding to the facial videos through a micro-expression recognition model. The assessment unit is used to perform phased dynamic assessments of emotional states based on multimodal features obtained by fusing brain activity emotional outcomes and emotional facial representations. The determination unit is used to determine the current emotional state of the air rifle athlete based on the results of the phased dynamic assessment and the preset emotional state classification table.

[0008] Thirdly, this application provides a processing device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor executes the method provided in the first aspect of this application when it invokes the computer program in the memory.

[0009] Fourthly, this application provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the method provided in the first aspect of this application.

[0010] From the above, it can be concluded that this application has the following beneficial effects: With the goal of monitoring the emotions of air rifle athletes during training, this application uses a phase entanglement collaborative network and a micro-expression recognition model to extract deep spatiotemporal features from EEG signals and facial micro-expressions, achieving deep fusion of context awareness. This enables the synchronous and accurate identification of the complex emotional states of air rifle athletes in high-pressure and rapidly changing training environments, thereby providing high-quality data support for the monitoring and intervention of the emotional states of air rifle athletes. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a training emotion monitoring method for air rifle athletes based on phase entanglement collaboration, as proposed in this application. Figure 2 This is an example illustration of how the athlete wearing the device is described in this application. Figure 3 This is a schematic diagram illustrating an example of the extraction of emotional results from brain activity in this application. Figure 4 This is a schematic diagram illustrating an example of extracting facial expressions of emotion in this application. Figure 5 This is a schematic diagram of a phase entanglement-based emotion monitoring device for air rifle athletes based on the present application. Figure 6 This is a schematic diagram of one type of processing equipment used in this application. Detailed Implementation

[0013] The technical solutions of 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 of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. The naming or numbering of steps appearing in this application does not imply that the steps in the method flow must be performed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical purpose, as long as the same or similar technical effect is achieved.

[0015] The module division described in this application is a logical division. In practical applications, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual coupling, direct coupling, or communication connections may be through interfaces, and the indirect coupling or communication connections between modules may be electrical or other similar forms, none of which are limited in this application. Moreover, the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed across multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in this application.

[0016] Before introducing the phase entanglement-based collaborative method for monitoring training emotions of air rifle athletes provided in this application, we will first introduce the background information involved in this application.

[0017] The method, device, and computer-readable storage medium for monitoring training emotions of air rifle athletes based on phase entanglement collaboration provided in this application can be applied to processing devices. Based on the deep spatiotemporal features extracted from EEG signals and facial micro-expressions by phase entanglement collaboration network and micro-expression recognition model, it realizes deep fusion of context awareness, thereby enabling synchronous and accurate identification of complex emotional states of air rifle athletes in high-pressure and rapidly changing training environments, and thus providing high-quality data support for monitoring and intervention of the emotional states of air rifle athletes.

[0018] The phase entanglement-based collaborative method for monitoring training emotions in air rifle athletes mentioned in this application can be implemented by a phase entanglement-based collaborative air rifle athlete training emotion monitoring device, or by different types of processing devices such as servers, physical hosts, or user equipment (UE) integrating the phase entanglement-based collaborative air rifle athlete training emotion monitoring device. The phase entanglement-based collaborative air rifle athlete training emotion monitoring device can be implemented in hardware or software. The UE can specifically be a terminal device such as a smartphone, tablet, laptop, desktop computer, or personal digital assistant (PDA). The processing devices can be configured in a device cluster.

[0019] It is understandable that the proposed solution is usually based on existing data or data that has already been collected. Therefore, the processing equipment that implements the phase entanglement-based collaborative air rifle athlete training emotion monitoring method of this application or that is equipped with the corresponding application service of the phase entanglement-based collaborative air rifle athlete training emotion monitoring method of this application usually only needs to meet the required data processing capabilities, and its specific equipment type and equipment deployment form are quite flexible.

[0020] If the direct acquisition of existing data mentioned above is also involved, then further hardware and software adaptations are needed for the processing equipment to enable it to acquire data. For example, if real-time acquisition of EEG signals and / or facial video is required, the corresponding acquisition equipment (i.e., EEG signal acquisition system and / or camera device) can be incorporated into the processing equipment cluster. Alternatively, the processing equipment itself can be the control unit for these acquisition devices. Or, a third-party call can be used to trigger external acquisition devices to perform real-time acquisition of EEG signals and / or facial video.

[0021] In addition, if there is a need to display the processing progress (including the processing results), the processing device itself can be configured with the required display screen (including touch screen) to display the specific content. Of course, the processing device can also display the specific content through an external display device or other devices with a display screen.

[0022] The following section introduces the phase entanglement-based collaborative method for monitoring training emotions in air rifle athletes, as provided in this application.

[0023] First, refer to Figure 1 , Figure 1 This paper illustrates a flowchart of a training emotion monitoring method for air rifle athletes based on phase entanglement collaboration, as described in this application. The training emotion monitoring method for air rifle athletes based on phase entanglement collaboration provided in this application may specifically include the following steps S101 to S104: Step S101: Obtain the EEG signals and facial video of the current air rifle athlete during training; Understandably, the processing involved in this application is based on the EEG signals and facial videos of current air rifle athletes during training.

[0024] The "current air rifle athlete" here does not refer to the current situation, but rather to the air rifle athlete targeted by the current solution. This is easy to understand, and in some cases, it may involve the need to monitor the air rifle athlete's training emotions during past sports activities.

[0025] In this context, the acquisition and processing of EEG signals and facial videos can be done in real time, with real-time and readily available data, or with past data. It can also involve corresponding manual configuration, which can be flexibly configured according to actual needs in practical applications.

[0026] Furthermore, in specific applications, the proposed solution is typically initiated as a training emotion monitoring task for air rifle athletes. This task can be initiated manually or by the system according to a corresponding autonomous initiation strategy. The EEG signals and facial videos to be acquired can be directly included in the task information, or the acquisition methods (including acquisition addresses) of these two data aspects can be included in the task information to serve as indirect instructions. Alternatively, other types of data acquisition methods can be used in conjunction with the task information, which is quite flexible.

[0027] As an example, EEG signals can be obtained using the wireless multichannel EEG acquisition system g.NautilusPRO, while facial videos can be obtained using miniature cameras.

[0028] Furthermore, it is understood that this application may also involve corresponding preprocessing to improve data quality. Specifically, preprocessing may be performed after the initial EEG signal and initial facial video are obtained here, or the EEG signal and facial video obtained here may have already been preprocessed, or the preprocessing may be integrated into the temporal alignment operation mentioned later, which is quite flexible.

[0029] As yet another example, on the one hand, the EEG signals acquired here can be denoted as... , , The number of EEG signal sampling points; the facial video acquired here can be denoted as... , .

[0030] Combination Figure 2 The diagram shown illustrates an example of an athlete wearing the device, where the athlete wears a wireless multi-channel EEG acquisition system (g.NautilusPRO) and a miniature camera to capture EEG signals and facial video.

[0031] Step S102: After performing temporal alignment on the EEG signals and facial videos, the brain activity and emotional results of the EEG signals are extracted through a phase entanglement collaborative network, and the emotional facial representations corresponding to the facial videos are extracted through a micro-expression recognition model. Understandably, after acquiring sensor data from two modalities, namely EEG signals and facial video, this application may also involve time alignment operations. For these two aspects of sensor data with time-series characteristics, the time sequence between them is aligned, thereby laying a better foundation for subsequent data processing.

[0032] After completing the temporal alignment, this application can perform corresponding feature extraction processing on the EEG signals and facial videos at that time through separately configured neural network models, in order to capture the emotional results of brain activity and emotional facial representations.

[0033] As can be understood, this involves feature extraction of two modalities, which are accomplished by different neural network models. The two can also form a fusion model architecture. This application can refer to the overall fusion model as the PheLLM model. Based on the powerful learning capabilities of the neural network model built by the machine learning methods of Artificial Intelligence (AI), efficient and high-performance feature extraction results can be obtained.

[0034] Furthermore, the neural network model for extracting brain activity and emotional results from EEG signals specifically employs the Phase Entanglement Collaborative Network (PhyNet) in this application. The micro-expression recognition model for extracting emotional facial representations from facial videos has a more flexible model type. As an example, the Multi-scale Emotion Recognition Network (MERNet) can be used.

[0035] Furthermore, for the micro-expression recognition model, the model input in specific operations can also be a more refined facial image obtained through further processing of facial videos, thereby enabling more delicate model processing. This facial image can be denoted as... T is the total number of sequence groups, and t is the identifier of the current sequence group.

[0036] Step S103: Based on the multimodal features obtained by fusing brain activity emotional results and emotional facial representations, conduct a phased dynamic assessment of emotional state. After extracting brain activity and emotional results and emotional facial expressions from different modalities using the two deep learning models mentioned above, it is understandable that a deep cross-modal feature fusion operation can be performed on the two to better represent the deep characteristics of current air rifle athletes during training.

[0037] Next, based on the obtained multimodal features, specific phased dynamic evaluations can be performed on the corresponding emotion state recognition target to obtain the required phased dynamic evaluation results.

[0038] It is understandable that this section can also involve the application of relevant machine learning methods, especially deep learning methods. Similar to the previous section, it aims to achieve efficient and high-performance phased dynamic evaluation results based on the powerful learning capabilities of AI algorithms.

[0039] Step S104: Determine the current emotional state of the air rifle athlete based on the phased dynamic assessment results and the preset emotional state classification table.

[0040] As can be seen, in the design of this application, the processing model does not directly output the specific emotional state of the current air rifle athlete. Instead, it provides an initial emotional state recognition result, which is then combined with the preset emotional state classification table designed here to finally output the emotional state recognition result.

[0041] In other words, the proposed solution does not output isolated emotion labels, but rather maps and correlates the emotion categories, intensities, and key evidence (model output) identified by the algorithm with a pre-defined athlete state emotion classification table. In practical applications, this design can give the output of technical indicators a clear meaning in sports psychology, greatly enhancing the interpretability and professionalism of the results, and making it easier to deploy more suitable intervention programs to achieve better adjustment of pre-exercise emotions.

[0042] from Figure 1 As can be seen from the embodiments shown, under the goal of monitoring the training emotions of air rifle athletes, this application, based on the deep spatiotemporal features extracted from EEG signals and facial micro-expressions by a phase entanglement collaborative network and a micro-expression recognition model, achieves deep fusion of context awareness. Thus, in a high-pressure, fast-changing training environment, it can synchronously and accurately identify the complex emotional state of air rifle athletes, thereby providing high-quality data support for the monitoring and intervention of the emotional state of air rifle athletes.

[0043] Continue with the above Figure 1 The steps of the illustrated embodiment and their possible implementation methods in practical applications are described in detail.

[0044] As an exemplary embodiment, the timing alignment operation involved in step S102 may specifically include the following processing content: 1.1) Input facial video Exporting as an image frame sequence at a fixed frame rate , A multi-task convolutional neural network is used to process image frame sequences. Each frame of the image undergoes face detection and key point alignment. Based on the processing results, five consecutive frames covering the complete dynamic process of micro-expressions are selected and uniformly scaled. Pixels are stacked to form the corresponding micro-expression feature tensor. , ; Among them, for image frame sequences Face detection in each frame of the image can be performed, as an example, using an SFE-DETR and SCRFD detector. The SFE-DETR and SCRFD detectors output the face detection results. , , For the first The center coordinates of the bounding box of a person's face. , , These are width, height, and detection confidence, respectively. This represents the total number of faces detected.

[0045] For image frame sequences Face detection for each frame in the dataset, as another example, can be performed using a multi-task cascaded convolutional neural network (MTCNN), with its output serving as the keypoint coordinates of the face detection results. , , For the first Two-dimensional coordinates of a person's face The coordinates of the five key points are: left eye, right eye, tip of nose, left corner of mouth, and right corner of mouth.

[0046] Next, key point alignment based on face detection results may involve the application of standard template points to align face images as much as possible, thereby improving data quality.

[0047] As an example, standard template points can be calculated to align key points in face detection results. coordinates of the detection point The similarity transformation parameters can be expressed as follows: , in, Scaling factor For rotation matrix, It is a translation vector. , The Umeyama algorithm is used for closed-form solution to ensure the accuracy and stability of face geometric normalization.

[0048] Next, from the aligned facial images, select 5 consecutive frames that can cover the complete dynamic process of micro-expressions. , And scale them uniformly to A pixel can be represented as follows: , , Then, they are stacked to form a micro-expression feature tensor. ,

[0049] 1.2) Micro-expression feature tensor For each video segment involved, a set of 5 frames is extracted at fixed time intervals to form a preliminary face image that is temporally aligned with the EEG signal dimension. T is the total number of sequence groups, and t is the identifier of the current sequence group; Understandable, for micro-expression feature tensors The newly constructed face video can then undergo frame image extraction. By extracting a set of 5 frames at fixed time intervals, preliminary temporally aligned face images are formed. , .

[0050] 1.3) Input EEG signals Bandpass filtering was applied in the range of 0.1Hz-45Hz, and a 1-second non-overlapping sliding window was used to divide the signal into EEG sequences that were initially time-aligned with the facial video dimensions. , C represents the number of channels in the EEG sequence. With facial images Forming time-aligned multimodal data pairs .

[0051] The signal partitioning operation can be represented as follows: , The alignment operation can be represented as follows: , in, For the first Timestamps of each EEG sampling point For the first The timestamp of the frame video, The maximum allowable synchronization deviation, In this way, ensure and Strict temporal correspondence forms time-aligned multimodal data pairs. This can be used for further model feature extraction.

[0052] As yet another exemplary embodiment, combined with Figure 3 The diagram shown is an example of the extraction of emotional results from brain activity in this application. For the Phase Entanglement Cooperative Network (PhyNet), the following processing may be included: 2.1) Using the temporal Transformer encoder module to extract EEG signals from the input network Extracting temporal dependency features The temporal Transformer encoder internally incorporates a hierarchical coding structure based on a phase-varying mode attention bias mechanism and a lightweight LLM modulator. The attention bias mechanism is used to perform initial screening of phase entanglement in the temporal dimension, while the lightweight LLM modulator is used to extract temporally dependent features based on the input context and the attention bias mechanism. ; Here, LLM stands for Large Language Model, which shows that this application involves LLM-based question-answering processing to obtain more refined coding features.

[0053] The temporal Transformer encoder is internally configured with a hierarchical coding structure based on a phase-varying time-varying mode attention bias mechanism and a lightweight LLM modulator, forming a hierarchical semantic modulation unit.

[0054] For example, the lightweight LLM modulator can operate according to the following standardized cognitive enhancement protocol: { "Prompt word system": { "Analysis Level": [ { "Hierarchy Name": "Temporal Pattern Resolution" "Cue word template": "Analyze the rhythmic fluctuation summary of the input neural temporal signal, identify potential event segments that significantly deviate from the baseline, and preliminarily determine which basic emotional dimension these events may be related to." "Output Target": "Generate a primary event heat vector" This vector labels the probability of significant neural events occurring at each time point and the rough correlation strength between them and the basic emotional dimension. }, { "Hierarchical Name": "Neighborhood Semantic Association" "Cue word template": "Based on the neural event patterns identified in the initial stage, combined with prior knowledge of training exercise psychology, these patterns are mapped to specific training psychological state segments, and the core time window that best represents the psychological state is precisely located." "Output Target": "Generate a refined semantic weight vector" "A mental state label and a weight vector precisely quantify the contribution of each time point to a specific training mental state." }, { "Level Name": "Modulation Signal Synthesis" "Prompt word template": "By combining the initial event popularity with refined semantic weights, the final time dimension importance distribution is synthesized, and high-confidence regions are enhanced while contradictory or low-confidence regions are suppressed, generating a smooth and highly discriminative modulation signal." "Output Target": "Final Temporal Importance Modulation Vector" This vector has incorporated multi-level analysis results and undergone confidence calibration, making it directly usable for feature modulation. }, "Execution Flow": "The three-layer analysis is executed sequentially, with the output of the previous layer serving as the input context for the next layer." "Output Specification": "Structured data output enables real-time feature focusing guided by semantics." } } As another example, this application also specifically designed a prompt word template to guide the multi-turn dialogue interaction process of the LLM model, with the following prompt word template examples: { "Dialogue Example": { "Interaction Sequence": [ { Round: 1 "User Query": "The intermediate feature summary of the current time-series coding shows that within the time segment [t-500ms, t-200ms], the signal energy of the prefrontal cortex channel group exhibits a sustained sharp increase in the β band (15-30Hz), and this pattern is detected in two consecutive coding blocks. Please analyze the basic properties of this neural pattern." "Model Response": "An event of 'abnormally increased power in the prefrontal β band' was detected in the athlete. This pattern is often associated with high-arousal negative emotional states such as 'increased cognitive load,' 'overthinking,' or 'anticipatory anxiety' in neuroscience. It is preliminarily identified as an emotion-related neural event that requires close attention." "Analysis Dimensions": ["Beta band energy anomaly", "Prefrontal cortex localization", "High arousal negative emotion association"] }, { Rounds: 2 "User Query": "Based on the 'prefrontal high beta wave' event identified in the previous round, and in conjunction with the 10-meter air rifle firing preparation phase, infer the training psychological state most likely mapped by this event, and assess its importance to the current emotional representation." “Model Response”: “Within the time window preceding the firing of a 10-meter pistol, high beta waves in the prefrontal cortex are a highly specific indicator of ‘anticipatory anxiety’ or ‘fear of failure’. This event makes a core contribution to the characterization of the ‘anxiety’ state at the current time point, and its importance is assessed as ‘high’.” "Analysis Dimension": ["Mapped to: Anticipatory Anxiety", "Task Contextual Relationship", "Core Contribution: High"] }, { Rounds: 3 User query: "Please synthesize the final modulated signal, with the following requirements: enhance the characteristic response of the time slice in which the event occurs, and ensure a smooth transition to avoid introducing noise." “Model Response”: “Generates a Gaussian-shaped weight distribution vector centered at time point t-350ms with a standard deviation of 100ms. This vector is assigned a peak weight of 0.9 in the interval [t-500ms, t-200ms] and decays smoothly towards both ends. This signal will be used to reweight the next layer of features or attention matrix to highlight the temporal features related to ‘anxiety’.” "Analysis Dimensions": ["Modulation Signal Shape: Gaussian Distribution", "Peak Weight: 0.9", "Target: Feature / Attention Reweighting"] } ], "Process Characteristics": { "Explicitly output-oriented": "The final output is not a text report, but a structured, executable mathematical vector that is directly integrated into the forward propagation of the neural network." "Goal-driven": "The interaction strictly serves the ultimate goal of 'generating temporally important modulation vectors', and each round is a necessary step in constructing the final modulation signal". "Data and knowledge fusion": "Inference simultaneously relies on real-time feature data (intermediate feature summaries) and domain prior knowledge internalized by LLM (neuro-psychological associations, sports psychology)". } } } The constructed complete input sequence is fed into the LLM, which extracts temporal dependency features based on this context. .

[0055] 2.2) In the constructed phase entanglement map Based on this, and combined with the phase-locked value calculation function, the EEG signal is analyzed. Spatial continuity features are extracted using the ODE module of the continuous graph. Among them, the phase entanglement map From the static phase entanglement matrix and dynamic phase entanglement matrix The resulting static phase entanglement matrix is ​​obtained through fusion. Dynamic phase entanglement matrix constructed based on phase lock value. Learning through adaptive mechanisms; It is understood that the following configuration conditions are involved in the extraction of brain activity and emotion results based on EEG signals in this application: 1. Physical basis: In multi-channel EEG signals, the instantaneous phases of different brain regions no longer remain statistically independent, but instead form a joint distribution structure with high-order dependencies; 2. Computational Representation: This high-order cooperative structure of phase relationship is modeled as a dynamic-static joint adjacency matrix of a graph neural network, where: Static entanglement component: constructed based on the phase locking index (PLI), representing a long-term stable brain functional connectivity framework; Dynamic entangled components: learned in real time through an adaptive mechanism, representing the instantaneous strain of phase relationships under task conditions; Entanglement fusion: Static entanglement components and dynamic entanglement components constitute a complete phase entanglement map.

[0056] 3. Evolutionary Mechanism: This phase entanglement map serves as the core operator for state transition in the Ordinary Differential Equation (ODE) module of the continuous graph, controlling the neural differential evolution process of EEG spatial features.

[0057] Specifically, for static phase entanglement matrix Based on the phase-locked value, it can be represented as follows: , For dynamic phase entanglement matrix Learning through an adaptive mechanism can be represented as follows: , in, , This is the weight matrix. , , The activation function is then fused into a dynamic-static graph. .

[0058] The Phase Lock Value (PLI) calculation can be expressed as follows: , Among them, the instantaneous phase is extracted by performing Hilbert transform on the EEG signal. Calculation channel With channel Phase lock value between.

[0059] The continuous graph ODE module extracts spatially continuous features. , can be represented as follows: , , in, This is the normalized graph adjacency matrix. For learnable parameters, These are the initial features.

[0060] 2.3) Temporal characteristics With spatial continuity The data are spliced ​​and linearly transformed to form a unified spatiotemporal representation modulated by a phase entanglement map, which serves as the output of emotional results of brain activity.

[0061] After obtaining the time series characteristics With spatial continuity After considering these two aspects, we can continue to splice and linearly transform them to form a unified spatiotemporal representation modulated by the phase entanglement spectrum, thus obtaining the brain activity and emotional results as the output.

[0062] This can be represented as follows: , Here, || represents the concatenation operation along the feature dimension.

[0063] Regarding the above settings, it is understandable that this application is the first to propose the high-order phase collaborative modeling paradigm of "phase entanglement" in the field of EEG emotion recognition, and has implemented it in engineering based on the PhyNet model. Compared with traditional brain functional connectivity networks based on phase-locked values, its substantial progress lies in: From "pairwise relationships" to "higher-order structures," it is possible to capture the coordinated change patterns of phase relationships across multiple brain regions; From “static snapshots” to “continuous evolution”: Embedding phase entanglement maps into the graph ODE framework makes the extraction of EEG spatial features no longer a stacking of discrete graph convolutions, but a neural differential evolution process driven by entanglement structures.

[0064] As yet another exemplary embodiment, combined with Figure 4 The illustration shown is an example of extracting emotional facial representations according to this application. The micro-expression recognition model is configured using MERNet, and the micro-expression recognition model includes the following processing: 3.1) Input the facial video into the model. Processed face images The data is segmented into spatiotemporal tiles and linearly projected, while learnable spatiotemporal location codes are added to form an initial token sequence. ; The input here is a face image. For specific processing details, please refer to the previous examples of timing alignment operations.

[0065] For which facial images The operation of uniformly dividing a spatiotemporal graph into N spatiotemporal tiles in both time and space can be represented as follows: , in, For time depth, For spatial height, For space width, The linear projection, or the operation of projecting each flattened tile onto a fixed token dimension D, can be represented as follows: , in, , For learnable parameters, Since the Transformer itself does not have the ability to perceive sequence order, unknown information can be added, as shown below: , in, Learnable spatiotemporal location encoding encodes the unique location of each tile in both the time series and two-dimensional space. .

[0066] 3.2) Initial token sequence Input a hierarchical spatiotemporal Transformer encoder to obtain high-level tokens. In the hierarchical spatiotemporal Transformer encoder, the shallow layer uses local window attention to capture subtle motions, while the deep layer uses global attention to understand cooperative patterns. At the same time, the adaptive importance weight module inside the encoder performs token dynamic weighting. Among them, the hierarchical spatiotemporal Transformer encoder, namely the mesTransformer encoder shown in the figure, uses local window attention and global attention in its shallow and deep layers, respectively. Each layer then includes multi-head attention, layer normalization, and a feedforward network.

[0067] It generates an importance score for each token through a lightweight network, which can be represented as follows: , And receive the token from the current layer. The resulting high-level token sequence can be represented as follows: , Shallow layers use local window attention to capture subtle movements, while deep layers use global attention to understand collaborative patterns, which can be represented as follows: , , Meanwhile, the internal adaptive importance weighting module dynamically weights the tokens.

[0068] 3.3) A set of learnable parameters will be introduced. As a query, with high-level tokens Perform cross-attention calculation to obtain the AU feature vector. ; in, ,Will As a query, the encoder output is used as both key and value, and cross-attention is applied to generate structured AU feature vectors with explicit physical meaning. The corresponding representation is as follows: , Where k is the structured feature vector of AU.

[0069] 3.4) For high-level tokens Aggregation yields a global spatiotemporal representation Perform pooling and combine with the AU feature vector By fusing the data, high-level semantic features are obtained. As an output of emotional facial representation.

[0070] Among them, for high-level tokens The aggregation yields a global, fixed-dimensional feature vector, i.e., a global spatiotemporal representation. , can be represented as follows: , and Gated fusion can be represented as follows: , Where g is the adaptive gating vector, the processing result can be represented as follows: .

[0071] Continuing with step S103, as an exemplary embodiment, the fusion of multimodal features and the phased dynamic assessment can be specifically performed based on the PheTransformer model. That is, based on the multimodal features obtained by fusing brain activity emotional results and emotional facial representations, the phased dynamic assessment of emotional states can include: Based on the multimodal features obtained by fusing brain activity emotional outcomes and emotional facial representations, the PheTransformer model is used to perform phased dynamic assessments of emotional states.

[0072] Understandably, this application, in addition to extracting deep spatiotemporal features from EEG signals and facial micro-expressions, also innovatively introduces an adaptive attention modulation graph mechanism based on PheTransformer. This mechanism can dynamically learn and weight the correlation strength between different modal features, achieving context-aware fusion. This enables better synchronous and accurate recognition of complex emotions in high-pressure, fast-changing training environments.

[0073] In this case, the PheTransformer model can specifically include the following processing: 4.1) The brain activity emotional outcomes and emotional facial representations are spliced ​​together and embedded in location encoding. Generate an initial token sequence representation ; This can be represented as follows: , in, For position encoding, , Then, location encoding information is added to preserve the spatiotemporal structure.

[0074] 4.2) In each layer of the network, combine the feature representation of the current layer. and guiding vector Dynamically calculate the attention modulation map that reflects the correlation strength between multimodal features. ; This can be represented as follows: .

[0075] 4.3) In the multi-head attention mechanism of windows, the attention modulation graph is... It is integrated into the attention weight calculation in the form of element-wise multiplication to explicitly adjust the interaction strength between different features, so as to realize the cross-window propagation of modulation information and global context modeling, and refine and enhance feature representation layer by layer; This involves attention modulation maps Integrating element-wise multiplication into attention weight calculation can be represented as follows: , in, As input features, For activation function, Explicitly adjusting the interaction strength between different features, the window attention involved can be represented as follows: , The alternation of attention with shifted windows can be represented as follows: , in, For window size, Circular shift operation, To achieve cross-window propagation of modulation information and global context modeling, and to refine and enhance feature representations layer by layer, it can be represented as follows: , in, To retain the coefficient, This is a lightweight update function.

[0076] 4.4) After processing by multiple Transformer blocks, the high-level features of the final output are globally pooled to convert them into feature vectors of fixed dimensions. Then, they are further processed by a classifier to obtain the specific emotional state classification probability, which is output as the stage dynamic evaluation result.

[0077] This can be represented as follows: , in, For the first The logit value for each category.

[0078] After completing the phased dynamic assessment of emotional state, the process can be mapped and correlated with the preset athlete emotional state classification table in step S104 to determine the final emotional state.

[0079] As an exemplary embodiment, the potential emotional states that may be involved in the phased dynamic evaluation results of this application specifically include four specific emotional states: anxiety, focus, excitement, and fatigue.

[0080] Building upon this, the preceding multimodal feature fusion and attention modulation aim to identify the phased emotional categories (such as anxiety, focus, excitement, and fatigue) exhibited by athletes during the assessment period, which can be represented as follows: , , in, , These represent micro-expression features and electroencephalogram (EEG) signal features, respectively. The responses are anxiety, focus, excitement, and fatigue.

[0081] At this point, the various emotion categories can be weighted and integrated, and the specific emotional state of the athlete during that period can be determined based on the judgment threshold in the preset training state emotion classification table.

[0082] As an example, the athlete emotional state classification table can be represented as follows: Table 1 - Example of an Athlete's Emotional State Classification Table

[0083] Once the final emotional state is determined, it is clear that data-driven decision support can be provided to determine whether athletes need to adjust their emotions. This can include outputting results, pushing results, storing them locally, storing them off-site, or performing further data analysis. The specific output processing can be flexibly adjusted according to pre-configured and real-time output strategies.

[0084] As an exemplary embodiment, for further data analysis and processing, after step S104, the method of this application may further include: Based on the emotional state, determine the outcome and identify the appropriate intervention plan for the current air rifle athlete.

[0085] The adaptive intervention plan can be directly carried or included in the previously mentioned athlete emotional state classification table, or it can involve matching with intervention plans from other places, or it can involve real-time generation of further intervention plans. In practical applications, this can be flexibly configured according to specific needs.

[0086] The above is an introduction to the air rifle athlete training emotion monitoring method based on phase entanglement collaboration provided in this application. In order to facilitate better implementation of the air rifle athlete training emotion monitoring method based on phase entanglement collaboration provided in this application, this application also provides an air rifle athlete training emotion monitoring device based on phase entanglement collaboration from the perspective of functional modules.

[0087] See Figure 5 , Figure 5 This is a schematic diagram of a phase entanglement-based collaborative air rifle athlete training emotion monitoring device according to this application. In this application, the phase entanglement-based collaborative air rifle athlete training emotion monitoring device 500 may specifically include the following structure: The acquisition unit 501 is used to acquire the electroencephalogram (EEG) signals and facial videos of the current air rifle athlete during training. The extraction unit 502 is used to extract the brain activity and emotional results corresponding to the EEG signal through a phase entanglement collaborative network after performing a temporal alignment operation on the EEG signal and the facial video, and to extract the emotional facial representation corresponding to the facial video through a micro-expression recognition model. Assessment unit 503 is used to perform phased dynamic assessment of emotional state based on multimodal features obtained by fusing brain activity emotional outcomes and emotional facial representations. Unit 504 is used to determine the current emotional state of the air rifle athlete based on the results of the phased dynamic assessment and the preset emotional state classification table.

[0088] In one exemplary embodiment, the timing alignment operation includes the following processing: Input facial video Exporting as an image frame sequence at a fixed frame rate , A multi-task convolutional neural network is used to process image frame sequences. Each frame of the image undergoes face detection and key point alignment. Based on the processing results, five consecutive frames covering the complete dynamic process of micro-expressions are selected and uniformly scaled. Pixels are stacked to form the corresponding micro-expression feature tensor. , ; micro-expression feature tensor For each video segment involved, a set of 5 frames is extracted at fixed time intervals to form a preliminary face image that is temporally aligned with the EEG signal dimension. T is the total number of sequence groups, and t is the identifier of the current sequence group; For input EEG signals Bandpass filtering was applied in the range of 0.1Hz-45Hz, and a 1-second non-overlapping sliding window was used to divide the signal into EEG sequences that were initially time-aligned with the facial video dimensions. , C represents the number of channels in the EEG sequence. With facial images Forming time-aligned multimodal data pairs .

[0089] In yet another exemplary embodiment, for a phase-entangled cooperative network, the following processing is included: The temporal Transformer encoder module was used to extract EEG signals from the input network. Extracting temporal dependency features The temporal Transformer encoder internally incorporates a hierarchical coding structure based on a phase-varying mode attention bias mechanism and a lightweight LLM modulator. The attention bias mechanism is used to perform initial screening of phase entanglement in the temporal dimension, while the lightweight LLM modulator is used to extract temporally dependent features based on the input context and the attention bias mechanism. ; In the constructed phase entanglement map Based on this, and combined with the phase-locked value calculation function, the EEG signal is analyzed. Spatial continuity features are extracted using the ODE module of the continuous graph. Among them, the phase entanglement map From the static phase entanglement matrix and dynamic phase entanglement matrix The resulting static phase entanglement matrix is ​​obtained through fusion. Dynamic phase entanglement matrix constructed based on phase lock value. Learning through adaptive mechanisms; Time series features With spatial continuity The data are spliced ​​and linearly transformed to form a unified spatiotemporal representation modulated by a phase entanglement map, which serves as the output of emotional results of brain activity.

[0090] In yet another exemplary embodiment, the micro-expression recognition model is configured using MERNet, and the micro-expression recognition model includes the following processing: Input model, composed of facial video Processed face images The data is segmented into spatiotemporal tiles and linearly projected, while learnable spatiotemporal location codes are added to form an initial token sequence. ; Initial token sequence Input a hierarchical spatiotemporal Transformer encoder to obtain high-level tokens. In the hierarchical spatiotemporal Transformer encoder, the shallow layer uses local window attention to capture subtle motions, while the deep layer uses global attention to understand cooperative patterns. At the same time, the adaptive importance weight module inside the encoder performs token dynamic weighting. A set of learnable parameters will be introduced. As a query, with high-level tokens Perform cross-attention calculation to obtain the AU feature vector. ; High-level tokens Aggregation yields a global spatiotemporal representation Perform pooling and combine with the AU feature vector By fusing the data, high-level semantic features are obtained. As an output of emotional facial representation.

[0091] In yet another exemplary embodiment, the fusion of multimodal features and the staged dynamic evaluation are performed based on the PheTransformer model, which includes the following processing: The brain activity's emotional outcomes and emotional facial representations are spliced ​​together and embedded with location encoding. Generate an initial token sequence representation ; In each layer of the network, the feature representation of the current layer is combined. and guiding vector Dynamically calculate the attention modulation map that reflects the correlation strength between multimodal features. ; In the multi-head attention mechanism of windows, the attention modulation graph is used. It is integrated into the attention weight calculation in the form of element-wise multiplication to explicitly adjust the interaction strength between different features, so as to realize the cross-window propagation of modulation information and global context modeling, and refine and enhance feature representation layer by layer; After processing by multiple Transformer blocks, the high-level features of the final output are globally pooled to convert them into fixed-dimensional feature vectors. Then, they are further processed by a classifier to obtain the specific emotional state classification probability, which is output as the stage-by-stage dynamic evaluation result.

[0092] In yet another exemplary embodiment, the potential emotional states involved in the phased dynamic assessment results include anxiety, focus, excitement, and fatigue.

[0093] In yet another exemplary embodiment, the determining unit 504 is further configured to determine the result based on the emotional state, determine the intervention plan suitable for the current air rifle athlete, and output it through the output unit 505.

[0094] This application also provides a processing device from a hardware architecture perspective. As mentioned earlier, in practice, a processing device may exist as a device cluster. In this case, each device in the device cluster can also be referred to as a processing device. See [reference needed]. Figure 6 , Figure 6 This diagram illustrates a structural schematic of the processing device of this application. Specifically, the processing device may include a processor 601, a memory 602, and an input / output device 603. The processor 601 executes the computer program stored in the memory 602 to implement, for example... Figure 1 The corresponding embodiments include the steps of the air rifle athlete training emotion monitoring method based on phase entanglement cooperation in the embodiments; or, when the processor 601 executes the computer program stored in the memory 602, it implements the following: Figure 5 Corresponding to the functions of each unit in the embodiment, the memory 602 is used to store the functions executed by the processor 601 as described above. Figure 1 The corresponding embodiment includes the computer program required for the phase entanglement-based collaborative method for monitoring the training emotions of air rifle athletes.

[0095] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 602 and executed by processor 601 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a computer device.

[0096] The processing device may include, but is not limited to, processor 601, memory 602, and input / output device 603. Those skilled in the art will understand that the illustrations are merely examples of the processing device and do not constitute a limitation on the processing device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the processing device may also include network access devices, buses, etc., and processor 601, memory 602, input / output device 603, etc., are connected via a bus.

[0097] Processor 601 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the processing device, connecting various parts of the device through various interfaces and lines.

[0098] The memory 602 can be used to store computer programs and / or modules. The processor 601 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 602 and by calling data stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created according to the use of the processing device, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0099] When processor 601 executes a computer program stored in memory 602, it can specifically perform the following functions: Acquire EEG signals and facial videos of current air rifle athletes during training; After temporally aligning EEG signals and facial videos, the brain activity and emotional results corresponding to the EEG signals are extracted through a phase entanglement collaborative network, and the emotional facial representations corresponding to the facial videos are extracted through a micro-expression recognition model. Based on the multimodal features obtained by fusing brain activity emotional outcomes and emotional facial representations, a phased dynamic assessment of emotional state is conducted. Based on the phased dynamic assessment results and the pre-set emotional state classification table, the current emotional state of the air rifle athlete is determined.

[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the phase entanglement-based air rifle athlete training emotion monitoring device, processing equipment, and its corresponding units described above can be found in the following reference: Figure 1 The description of the phase entanglement-based collaborative method for monitoring training emotions of air rifle athletes in the corresponding embodiment will not be repeated here.

[0101] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0102] Therefore, this application provides a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute the present application. Figure 1 The steps of the air rifle athlete training emotion monitoring method based on phase entanglement collaboration in the corresponding embodiment can be referred to as follows for specific operations. Figure 1 The description of the phase entanglement-based collaborative method for monitoring training emotions of air rifle athletes in the corresponding embodiments will not be repeated here.

[0103] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0104] Because of the instructions stored in the computer-readable storage medium, the present application can be executed as described above. Figure 1 The steps of the air rifle athlete training emotion monitoring method based on phase entanglement cooperation in the corresponding embodiment can therefore achieve the results of this application. Figure 1 The beneficial effects of the phase entanglement-based collaborative method for monitoring the training emotions of air rifle athletes in the corresponding embodiments are detailed in the preceding description and will not be repeated here.

[0105] The foregoing has provided a detailed description of the method, apparatus, processing device, and computer-readable storage medium for monitoring training emotions of air rifle athletes based on phase entanglement collaboration, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the core ideas of this application; furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for monitoring training emotions in air rifle athletes based on phase entanglement collaboration, characterized in that, The method includes: Acquire EEG signals and facial videos of current air rifle athletes during training; After performing a temporal alignment operation on the EEG signal and the facial video, the brain activity and emotional results corresponding to the EEG signal are extracted through a phase entanglement collaborative network, and the emotional facial representations corresponding to the facial video are extracted through a micro-expression recognition model. Based on the multimodal features obtained by fusing the brain activity emotional results and the emotional facial representation, a phased dynamic assessment of emotional state is performed. The current emotional state of the air rifle athlete is determined based on the phased dynamic evaluation results and the preset emotional state classification table.

2. The method according to claim 1, characterized in that, The timing alignment operation includes the following processing steps: Input facial video Exporting as an image frame sequence at a fixed frame rate , A multi-task convolutional neural network is used to process image frame sequences. Each frame of the image undergoes face detection and key point alignment. Based on the processing results, five consecutive frames covering the complete dynamic process of micro-expressions are selected and uniformly scaled. Pixels are stacked to form the corresponding micro-expression feature tensor. , ; For the micro-expression feature tensor For each video segment involved, a set of 5 frames is extracted at fixed time intervals to form a preliminary face image that is temporally aligned with the EEG signal dimension. T is the total number of sequence groups, and t is the identifier of the current sequence group; For input EEG signals Bandpass filtering was applied in the range of 0.1Hz-45Hz, and a 1-second non-overlapping sliding window was used to divide the signal into EEG sequences that were initially time-aligned with the facial video dimensions. , C represents the number of channels, and the EEG sequence With the face image Forming time-aligned multimodal data pairs .

3. The method according to claim 1, characterized in that, For phase-entangled cooperative networks, the following processing is included: The temporal Transformer encoder module was used to extract EEG signals from the input network. Extracting temporal dependency features The temporal Transformer encoder internally incorporates a hierarchical coding structure based on a phase-varying mode attention bias mechanism and a lightweight LLM modulator. The attention bias mechanism is used to perform initial phase entanglement screening in the temporal dimension, and the lightweight LLM modulator is used to extract the temporal-dependent features based on the input context and the attention bias mechanism. ; In the constructed phase entanglement map Based on this, and combined with the phase-locked value calculation function, the EEG signal is processed. Spatial continuity features are extracted using the ODE module of the continuous graph. The phase entanglement map From the static phase entanglement matrix and dynamic phase entanglement matrix The static phase entanglement matrix is ​​obtained by fusion. The dynamic phase entanglement matrix is ​​constructed based on the phase-locked value. Learning through adaptive mechanisms; The time series features With the aforementioned spatial continuity feature The data is spliced ​​and linearly transformed to form a unified spatiotemporal representation modulated by a phase entanglement map, which serves as the output of the emotional results of the brain activity.

4. The method according to claim 1, characterized in that, The micro-expression recognition model is configured using MERNet, and includes the following processing steps: Input model, composed of facial video Processed face images The data is segmented into spatiotemporal tiles and linearly projected, while learnable spatiotemporal location codes are added to form an initial token sequence. ; Initial token sequence Input a hierarchical spatiotemporal Transformer encoder to obtain high-level tokens. In the hierarchical spatiotemporal Transformer encoder, the shallow layer uses local window attention to capture subtle motions, the deep layer uses global attention to understand cooperative patterns, and the adaptive importance weight module inside the encoder performs token dynamic weighting. A set of learnable parameters will be introduced. As a query, with the aforementioned high-level token Perform cross-attention calculation to obtain the AU feature vector. ; For the high-level token Aggregation yields a global spatiotemporal representation Perform pooling and combine it with the AU feature vector. By fusing the data, high-level semantic features are obtained. This serves as the output of the emotional facial representation.

5. The method according to claim 1, characterized in that, The fusion and phased dynamic evaluation of the multimodal features are performed based on the PheTransformer model, which includes the following processing: The brain activity emotional results and the emotional facial representations are spliced ​​together and embedded with location encoding. Generate an initial token sequence representation ; In each layer of the network, the feature representation of the current layer is combined. and guiding vector Dynamically calculate the attention modulation map that reflects the correlation strength between multimodal features. ; In the multi-head attention mechanism of the window, the attention modulation map is... It is integrated into the attention weight calculation in the form of element-wise multiplication to explicitly adjust the interaction strength between different features, so as to realize the cross-window propagation of modulation information and global context modeling, and refine and enhance feature representation layer by layer; After processing by multiple Transformer blocks, the high-level features of the final output are globally pooled to convert them into feature vectors of fixed dimensions. Then, they are further processed by a classifier to obtain the specific emotional state classification probability, which is output as the stage-based dynamic evaluation result.

6. The method according to claim 1, characterized in that, The potential emotional states involved in the phased dynamic assessment results include anxiety, focus, excitement, and fatigue.

7. The method according to claim 1, characterized in that, The method further includes: Based on the emotional state, determine the results, identify the appropriate intervention plan for the current air rifle athlete, and output it.

8. A training emotion monitoring device for air rifle athletes based on phase entanglement cooperation, characterized in that, The device includes: The acquisition unit is used to acquire the electroencephalogram (EEG) signals and facial videos of the current air rifle athlete during training. The extraction unit is used to extract the brain activity and emotional results corresponding to the EEG signal through a phase entanglement collaborative network after performing a temporal alignment operation on the EEG signal and the facial video, and to extract the emotional facial representation corresponding to the facial video through a micro-expression recognition model. An evaluation unit is used to perform a phased dynamic evaluation of emotional state based on the multimodal features obtained by fusing the emotional results of the brain activity and the emotional facial representation. The determining unit is used to determine the current emotional state of the air rifle athlete based on the phased dynamic evaluation results and the preset emotional state classification table.

9. A processing device, characterized in that, The method includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the method as described in any one of claims 1 to 7 when it invokes the computer program in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the method of any one of claims 1 to 7.

Images