Emotion intervention method based on wearable BCI and interactive hybrid augmented reality

Abstract

The invention discloses an emotion intervention method based on a wearable BCI and interactive hybrid augmented reality, and the method achieves the expansion and on-demand generation of a sand tool library through the construction of an interactive personalized virtual reality sand table environment and the combination of AIGC, and improves the interaction smoothness, personalized expression capability and immersion experience. Meanwhile, electroencephalogram signals are collected and analyzed in real time, the emotional state of the user is objectively and continuously evaluated, and therefore active interaction and objective monitoring in the emotional intervention process are combined. The method can support instant state evaluation of single intervention and long-term effect evaluation of multiple interventions at the same time, has good application value in the aspects of emotion regulation, psychological function improvement and intervention process tracking, and can reduce dependence on manual observation and subjective reports; and the participation degree, continuity and practical application value of intervention can be improved.

Description

Technical Field

[0001] This invention belongs to the field of biosignal processing and psychological intervention technology, specifically relating to an emotion intervention method based on wearable BCI (brain-computer interface) and interactive mixed extended reality. Background Technology

[0002] Emotional states have a significant impact on human mental health and are closely related to affective disorders such as depression, anxiety, and fatigue. These mental health disorders not only affect people's daily lives and work performance but also impose a significant social burden, highlighting the importance of developing effective and scalable emotion regulation interventions. Existing emotion interventions can be broadly categorized into pharmacological interventions, neuromodulation interventions, and behavioral interventions. Pharmacological interventions primarily regulate emotions by influencing neurochemical processes in the brain, but may have side effects such as dependence, headaches, and dry mouth. Neuromodulation interventions (such as electrical and magnetic stimulation) regulate neural activity by stimulating specific nerve sites, but are typically confined to medical settings and may raise safety concerns, making them difficult to use in daily life. In contrast, behavioral interventions alter an individual's emotional and behavioral responses through structured psychological strategies, making them safer and less costly than other intervention methods. However, many behavioral interventions still rely on professionals conducting them offline, lacking convenience for daily use and widespread application. Therefore, there is an urgent need for portable and wearable behavioral intervention technologies.

[0003] With the development of digital technology, mixed extended reality (MAR)-based interventions are gradually becoming a promising form of behavioral intervention. Virtual reality environments offer advantages such as immersion, controllability, and repeatability, facilitating standardized intervention processes, reducing reliance on specific intervention locations and mental health professionals, and improving intervention convenience. However, most existing virtual reality interventions primarily focus on passive content presentation with limited user interaction, which can easily lead to decreased engagement and sustained commitment over long-term use. To enhance user engagement, increasing research is focusing on virtual reality interventions that support active interaction. Sandplay, as an expressive and reflective form of psychological intervention, has been introduced into virtual reality scenarios. By constructing sandplay scenes in a virtual environment, people can externalize their inner thoughts into editable scene elements and iteratively express and reflect through continuous adjustment and reconstruction, thereby supporting emotion regulation. However, existing virtual reality sandplay intervention technologies have key limitations, including the following: Limitations on interactivity: Existing virtual reality sandbox systems typically have limited libraries of miniature models due to factors such as copyright, storage, cost, and manual selection and maintenance. These libraries are small in scale and struggle to support rich symbolic expression. When users cannot find suitable objects during scene construction, the expression process is easily interrupted, disrupting the continuity of interaction. Furthermore, some existing systems lack efficient and adjustable operation methods, resulting in heavy operational burdens and cumbersome interactions, thus weakening sustained engagement and immersive experience.

[0004] Personalization limitations: The scene elements in existing virtual reality sandboxes are mostly predefined content, and a large number of sandbox objects are directly assembled from general off-the-shelf resources, lacking customized design for individual expression needs; these objects often carry preset semantics, which do not completely match the story and imagery that users want to express, and users can only make approximate expressions, resulting in insufficient personalized expression ability and limited narrative details, thus affecting the intervention effect.

[0005] Limitations of lack of objective and real-time assessment: In existing virtual reality sandplay intervention studies, emotional state assessments typically rely on therapist observations and participant subjective reports. This approach lacks unified objective standards, and assessment results are easily influenced by subjective judgments, resulting in weak comparability across time periods and individuals. Furthermore, such assessments are often collected only at a limited number of time points, and analyses are usually limited to comparisons before and after the intervention, making it difficult to capture the continuously changing emotional dynamics throughout the intervention process. Summary of the Invention

[0006] In view of the above, the present invention provides an emotion intervention method based on wearable BCI and interactive hybrid extended reality, which can reduce the reliance on human observation and subjective reports, and help improve the participation, sustainability and practical application value of the intervention.

[0007] An emotion intervention method based on wearable BCI and interactive mixed extended reality includes the following steps: (1) Construct an interactive virtual reality sand table intervention scene, complete the three-dimensional modeling of the sand table space, configure the sand table and sand tool display area, and preset the interactive operation elements for scene construction; (2) Based on AIGC (Artificial Intelligence Generated Content) technology, a scalable 3D sandplay tool library is built to generate personalized sandplay tools on demand; (3) The EEG (electroencephalogram) and fNIRS (functional near-infrared spectroscopy) signals of the subjects were collected simultaneously through wearable BCI devices throughout the intervention process, and the head motion inertial signal was collected simultaneously as a reference to remove motion artifacts in the EEG and fNIRS signals. (4) Establish a structured emotional intervention paradigm to collect multimodal data, including questionnaire information, EEG and fNIRS signals, sand table scene layout, and interview information, at the corresponding stages before intervention, intervention, post-intervention and follow-up. (5) The effect of the emotional intervention was comprehensively evaluated based on the multimodal data obtained during multiple interventions, and the emotional change trend and sandplay interaction style characteristics of the subjects during multiple interventions were analyzed.

[0008] Furthermore, the interactive virtual reality sand table intervention scene in step (1) is a virtual room scene immersively constructed using 3D modeling software. It has a central table and a rectangular sand table inside, the surface of which is covered with a layer of sand. The inner edge of the sand table is set in blue to simulate the "water boundary" in the physical sand table. Shelves for placing sand objects are set on both sides of the sand table so that the subject can select and place sand objects in the virtual reality environment. The constructed scene has a wealth of interactive operation elements, including: moving and rotating the viewpoint in a first-person perspective using a hand controller; adding or deleting water areas on the surface of the sand table using the brush and eraser tools (and the size of the brush and eraser's effective range is adjustable); selecting sand objects from the shelves and placing them in the sand table, and continuously adjusting the posture and scaling the size of the placed sand objects.

[0009] Furthermore, the personalized sand figures in step (2) are automatically generated by the AIGC model. The 3D sand figure library includes at least the following 11 basic categories: people, animals, vegetation, buildings, vehicles, fences and signs, natural objects, fantasy objects, spiritual and mystical objects, landscape components, and household items. The AIGC model generates prompts according to emotional design constraints, including: using soft and harmonious color schemes with moderate contrast; using clear outlines and reasonable proportions for shapes and appearances, avoiding sharp structures and maintaining consistent scale; maintaining a balanced and stable spatial structure; and maintaining a consistent overall style for visual semantics, avoiding violent or uncomfortable appearances. The generated prompt rules are also used for on-demand sandplay creation during the intervention process to achieve dynamic expansion of the sandplay library and personalized symbolic expression. For sandplay generated by AIGC, it will be screened based on structural integrity, appearance consistency and emotional adaptability, and sandplay with strange structure, visual abnormality or inappropriate emotional expression will be removed to form a three-dimensional sandplay library for intervention.

[0010] Furthermore, the wearable BCI device in step (3) is a forehead-attached wireless device, which includes an EEG acquisition module, an fNIRS acquisition module, and an inertial acquisition module. The EEG acquisition module and the fNIRS acquisition module are integrated in the same attachment structure in a stacked manner to achieve synchronous acquisition of EEG and fNIRS signals in the same area of ​​the forehead. The inertial acquisition module includes an accelerometer and a gyroscope, which are used to acquire head motion inertial signals and provide a reference for subsequent EEG and fNIRS signal motion artifact suppression.

[0011] Furthermore, in step (3), the EEG signal is acquired from four electrodes, Fpz, AF7, AF8 and M2, according to the international 10-20 system standard, with M1 as the reference electrode and a sampling rate of 250Hz; the fNIRS signal is acquired in the bilateral forehead region through eight optical probes, using dual wavelengths of 735nm and 850nm and a sampling rate of 25Hz; during the intervention, the head motion inertial signal of 90.91Hz from the virtual reality head-mounted display and 12.5Hz from the wearable BCI device are simultaneously acquired as reference signals to adaptively remove artifacts in the EEG signal.

[0012] Furthermore, the emotional intervention paradigm in step (4) includes four stages: pre-intervention, intervention, post-intervention, and follow-up. The intervention stage lasts for two weeks, twice a week, for a total of four intervention sessions. The themes are energy, travel, connection, and birth, respectively. Each intervention session includes theme guidance, 20 minutes of virtual reality sand table construction, and review and reflection. The review and reflection includes guided reflection and semi-structured interviews. The sand table scene layout and interview information will be saved after each intervention. Resting-state EEG and fNIRS signals of the subjects are collected in the pre-intervention stage, the post-intervention stage, and before and after each intervention. Task-state EEG and fNIRS signals of the subjects are continuously collected during the virtual reality sand table construction process. At the same time, the subjects' state questionnaire and trait questionnaire scale are collected. Interview and questionnaire information are collected again in the follow-up stage to evaluate the sustainability of the intervention effect.

[0013] Furthermore, the questionnaire information in step (4) includes two types: trait questionnaires and state questionnaires. The effectiveness of the emotional intervention is comprehensively evaluated based on the scores of each questionnaire. Trait questionnaires are collected before, after, and during the follow-up phases, including: The PHQ (Patient Health Questionnaire) and BDI (Baker Depression Scale) are used to assess the degree of depression; the higher the score, the more severe the depressive symptoms. The GAD (Generalized Anxiety Disorder Scale), BAI (Baker Anxiety Scale), and STAI-T (Trait Anxiety Subscale) are scales whose scores are used to assess anxiety levels; the higher the score, the more severe the anxiety. The CD-RISC (Psychological Resilience Scale) is a scale used to assess psychological resilience. A higher score indicates that an individual has a higher level of psychological resilience and resilience. The FS (Fatigue Scale) is a scale used to assess the degree of fatigue; a higher score indicates more severe fatigue. The SRSS (Self-Rating Sleep Scale) is a scale used to assess the severity of sleep problems; a higher score indicates a more pronounced sleep problem. The Attention Control Scale (ACS) is used to assess attention control ability; a higher score indicates stronger attention control ability. A status questionnaire was collected before and after each intervention session, including: SAM (Self-Assessment Model) scores are used to assess a subject’s current valence and arousal status. The NRS (Digital Rating Scale) assesses the transient intensity of feelings such as pleasure, happiness, calmness, relaxation, anger, disgust, fear, anxiety, sadness, and dizziness. A higher score indicates a higher intensity of the corresponding emotion or state. PANAS (Positive and Negative Affect Inventory) scores assess current levels of positive and negative emotions. A higher score on the positive emotion subscale indicates stronger positive emotions, and a higher score on the negative emotion subscale indicates stronger negative emotions. STAI-S (State Anxiety Subscale) scores assess the level of state anxiety; a higher score indicates a more severe level of current anxiety.

[0014] Further, in step (5), the acquired EEG and fNIRS signals need to be processed and EEG features extracted. Specifically: First, the first 10 seconds and the last 10 seconds of each signal segment are removed. The EEG signal is then subjected to bandpass filtering of 1~45Hz and divided into time window segments with a length of 4 seconds and an overlap rate of 50%. Noise segments are removed based on amplitude thresholds and standard deviation thresholds. Then, the processed EEG signal is used to extract parameters including kurtosis, HFD (Higuchi fractal dimension), detrended fluctuation analysis parameters, power of each frequency band and frequency band power ratio, and Hjorth factor. (Hosse) Features including complexity and mobility characteristics, sample entropy, spectral entropy, and differential entropy. In addition, based on the frequency band power of AF7 and AF8 channels, the asymmetric features of the EEG prefrontal cortex are calculated. For the fNIRS signal, the change in optical density (OD) is obtained by conversion and removal of head motion artifacts. Then, according to the modified Beer-Lambert law, it is converted into the change in hemoglobin concentration and bandpass filtered at 0.01~0.1Hz to obtain the characteristic signals of the time-varying concentrations of oxyhemoglobin (HbO), deoxyhemoglobin (HbR), and total hemoglobin (HbT).

[0015] Furthermore, the comprehensive evaluation of the emotional intervention effect in step (5) includes two methods: long-term pre- and post-intervention evaluation and single-intervention pre- and post-intervention evaluation. The long-term pre- and post-intervention evaluation method is used to analyze the cumulative effect of multiple interventions. By comparing the questionnaire information and resting-state EEG characteristic signals at the initial stage of the intervention with those after several interventions, the overall trend of the subject's emotional state is evaluated. The single-intervention pre- and post-intervention evaluation method is used to analyze the changes in the subject's state before and after a single sandplay intervention. By comparing the questionnaire information collected before and after the intervention and the resting-state and task-state EEG characteristic signals, the immediate impact of a single intervention on the subject's emotional level and related physiological state is evaluated.

[0016] Furthermore, the analysis of sandplay interaction style features in step (5) is based on multimodal data to group and identify subjects. Specifically: First, the narrative text, sandplay scene layout information, EEG features, and questionnaire information of the subjects during the sandplay intervention process are extracted, and corresponding feature representations are constructed respectively. The narrative text is encoded into narrative semantic features through a text embedding model, the sandplay scene layout information is represented as layout features through scene spatial distribution and object operation descriptors, the EEG features are used to represent the neural activity state of the subjects during the intervention process, and the questionnaire information is represented by scale scores to represent the emotional state features of the subjects. Then, the dimensionality reduction of each modal data feature is performed, and the dimensionality reduction multimodal data features are spliced ​​and fused to form a joint feature representation for representing the sandplay interaction style of the subjects. Finally, based on the joint feature representation, a clustering algorithm is used to classify the subjects to distinguish the sandplay interaction style of the subjects.

[0017] A computer device includes a memory and a processor, wherein the memory stores a computer program and the processor executes the computer program to implement the aforementioned emotion intervention method based on wearable BCI and interactive mixed extended reality.

[0018] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned emotion intervention method based on wearable BCI and interactive mixed extended reality.

[0019] This invention constructs an interactive sandplay scene within an immersive virtual reality environment and combines AIGC technology to achieve large-scale expansion and on-demand personalized generation of a 3D sandplay tool library. This enhances the smoothness of interaction, the richness of symbolic resources, and the ability for individualized expression during sandplay intervention, overcoming the limitations of existing virtual reality sandplay interventions in terms of interactivity and personalization. Simultaneously, this invention introduces a wearable BCI device to simultaneously collect EEG and fNIRS signals during the intervention, enabling continuous and objective monitoring of the subject's emotional state. This reduces reliance on subjective reports and manual observation in traditional interventions, thereby enhancing the objectivity and comparability of the assessment results. Furthermore, this invention combines questionnaire data from before, during, and after the intervention, as well as EEG signals, interview records, and sandplay scene layout information during the intervention process, to conduct a multimodal comprehensive evaluation of the emotional intervention effect. This evaluation can simultaneously reflect the immediate changes from a single intervention and the cumulative effects of multiple interventions. Through the above technical solutions, this invention forms a complete technical path integrating active interaction, personalized expression, neuro-objective assessment, and long-term effect tracking. It has the characteristics of clear structure, strong scalability, and high feasibility of practical deployment. It can be used for remote health services related to emotion regulation and psychological intervention, and has good application prospects. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall framework of the emotion intervention method based on wearable BCI and interactive hybrid extended reality of the present invention.

[0021] Figure 2 This is a schematic diagram of a virtual sandbox scene in an embodiment of the present invention.

[0022] Figure 3 This is a flowchart illustrating the emotion intervention paradigm in this invention.

[0023] Figure 4 This is a schematic diagram showing the changes over time of multiple features of multichannel task-state EEG during a sandbox task in an embodiment of the present invention.

[0024] Figure 5 This is a schematic diagram illustrating the changes in deoxyhemoglobin concentration during the intervention period in an embodiment of the present invention. Detailed Implementation

[0025] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown, this embodiment provides an emotion intervention method based on wearable BCI and interactive mixed extended reality. The specific steps are as follows: (1) Set up an interactive virtual reality sand table intervention scene, complete the three-dimensional modeling of the sand table space, configure the sand table and sand tool display area, and preset the interactive operation elements for scene construction.

[0027] In this embodiment, the virtual reality sandbox intervention scene is modeled using 3D modeling software to immerse and construct a virtual room scene, such as... Figure 2 As shown, the system includes a central tabletop and a rectangular sandbox covered with sand. The inner edge of the sandbox is set in blue to simulate the "water boundary" in a physical sandbox. Shelves on both sides of the sandbox are provided for placing sand objects, allowing subjects to select and place objects in the virtual reality environment. The constructed scene includes a wealth of pre-defined interactive operations: first-person perspective viewpoint movement and rotation via a controller; adding or deleting water areas on the sandbox surface using brush and eraser tools, with adjustable range of influence; selecting sand objects from the shelves and placing them in the sandbox, with continuous posture adjustment and size scaling of the placed objects.

[0028] The sand table scene modeling program includes an interactive control module and a communication monitoring module. The interactive control module enables scene editing operations during sand table intervention, including selecting, placing, rotating, and scaling sand table objects, as well as drawing and erasing water areas on the sand table surface. The communication monitoring module continuously receives data or commands from external terminals through a preset interface within the virtual reality system, parses them, and triggers corresponding scene update operations based on the parsing results. The program can communicate with a remote server to load, display, and subsequently control personalized sand table generation results.

[0029] (2) Based on AIGC technology, an scalable 3D sandplay library can be built, which can generate personalized sandplay on demand.

[0030] This implementation utilizes AIGC technology to generate various 3D sandplay figures based on text prompts and preset aesthetic rules. These figures fall into 11 basic categories: people, animals, vegetation, buildings, vehicles, fences and signs, natural objects, fantasy objects, spiritual and mystical items, landscape components, and household items. The generation rules adhere to emotional design constraints, including: using soft and harmonious color schemes with moderate contrast; employing clear outlines, reasonable proportions, avoiding sharp structures, and maintaining consistent scale; maintaining a balanced and stable spatial structure; and ensuring consistent overall visual style, avoiding violent or disturbing appearances. These prompts are also used for on-demand sandplay figure generation during the intervention process, enabling dynamic expansion of the sandplay figure library and personalized symbolic expression. The AIGC-generated 3D sandplay figures are screened based on structural integrity, appearance consistency, and emotional suitability, eliminating candidate figures with bizarre structures, visual anomalies, or inappropriate emotional expressions to form a sandplay figure library suitable for emotional intervention.

[0031] Based on this, this implementation method adopts a multi-threaded or asynchronous processing mechanism to ensure that the sand table creation process does not block the current interactive flow. During the intervention, users can interact through the virtual interface, input prompts, and customize the generation of new 3D sand table as needed. The sand table creation process does not block the current interactive flow, thus enabling the sand table construction to continue while new sand table is being generated.

[0032] (3) The EEG and fNIRS signals of the subjects were collected synchronously through wearable BCI devices during the entire intervention process, and the head motion inertial signal was collected synchronously as a reference to remove motion artifacts in the EEG signal.

[0033] In this embodiment, a forehead-attached wearable BCI device is used to acquire EEG signals. This BCI device is a wireless device. The EEG acquisition module and the fNIRS acquisition module are integrated in the same attachment structure in a stacked manner to achieve synchronous acquisition of EEG and fNIRS signals in the same forehead region. The inertial acquisition module includes an accelerometer and a gyroscope to acquire head movement signals and provide a reference for subsequent brain signal motion artifact suppression. EEG signals are acquired from Fpz, AF7, AF8 and M2 electrodes according to the international 10-20 system standard, with M1 as the reference electrode, at a sampling rate of 250Hz. fNIRS signals are acquired in the bilateral forehead regions through eight optical probes, using dual wavelengths of 735nm and 850nm, at a sampling rate of 25Hz. In addition, during the intervention, head movement inertial signals of 90.91Hz from the virtual reality headset and 12.5Hz from the wearable BCI device are simultaneously acquired as reference signals to remove artifacts in the EEG signals.

[0034] After the wearable BCI device acquires signals in real time, the signals acquired by the patch-type BCI device are sent to the terminal device via Bluetooth communication for storage and analysis of EEG signals, thereby realizing cross-terminal data interaction between the EEG analysis terminal and the immersive virtual reality device. Motion artifact removal is performed on the terminal device using a Normalized Least Mean Square (NLMS) adaptive filtering algorithm to suppress motion artifacts. First, noise is estimated using the current weights, then the error between the real EEG and the noise is calculated, and this error is introduced... To prevent division by zero, gradient descent is used to minimize... The weights are then iteratively updated as follows: in: n Indicates the current moment. It is the noise reference signal used as the reference input vector. It is the adaptive filter weight vector at the current moment. These are the estimated artifacts and motion noise signals. It refers to the input brainwave signals. It is the estimated value of the EEG signal after artifact removal. The base step size is used to control the speed of weight updates. It is a standardized step size that adaptively adjusts its size based on the current reference signal.

[0035] (4) Set up a structured emotional intervention paradigm and collect data such as questionnaires, brain signals, sandplay scene layout and user interviews at the corresponding stages before, during, after and during the intervention.

[0036] The structured intervention paradigm in this implementation includes pre-intervention, intervention, post-intervention, and follow-up phases; such as... Figure 3 As shown, the intervention phase lasted for two weeks, twice a week, for a total of four intervention sessions, with the themes of energy, travel, connection, and birth in sequence. Each intervention session included a theme-guided session, a 20-minute virtual reality sand table construction, and a review and reflection session. The review and reflection included guided reflection and semi-structured interviews. After each intervention, the sand table layout and interview information were saved. Resting-state EEG and fNIRS signals were collected before and after the intervention, as well as before and after each intervention session. Task-oriented EEG and fNIRS signals were continuously collected during the virtual reality sand table construction. Status questionnaires and trait questionnaires were also collected. Interviews and questionnaires were collected again during the follow-up phase to assess the sustainability of the intervention effect.

[0037] The questionnaire information collected in this implementation includes two types: trait questionnaires and state questionnaires. The effectiveness of the emotional intervention is comprehensively evaluated based on the scores of each scale. Trait questionnaires were collected before, after, and during the follow-up phases. These included: PHQ and BDI scale scores to assess the degree of depression (higher scores indicate more severe depressive symptoms); GAD, BAI, and STAI-T scale scores to assess anxiety levels (higher scores indicate more severe anxiety); CD-RISC scale scores to assess psychological resilience (higher scores indicate greater psychological resilience and recovery ability); FS scale scores to assess the degree of fatigue (higher scores indicate more severe fatigue); SRSS scale scores to assess the degree of sleep problems (higher scores indicate more pronounced sleep problems); and ACS scale scores to assess attentional control ability (higher scores indicate stronger attentional control ability). State questionnaires were collected before and after each intervention session, including: SAM scale scores to assess the subject's current valence and arousal state; NRS scale scores to assess the transient intensity of pleasure, happiness, calmness, relaxation, anger, disgust, fear, anxiety, sadness, and dizziness, with higher scores indicating higher intensity of the corresponding emotion or state; PANAS scale scores to assess the current level of positive and negative emotions, with higher scores on the positive emotion subscale indicating stronger positive emotions and higher scores on the negative emotion subscale indicating stronger negative emotions; and STAI-S scale scores to assess the level of state anxiety, with higher scores indicating more severe current anxiety.

[0038] (5) Based on the questionnaire data, EEG and fNIRS signals, interview records and sandplay scene layout obtained during multiple interventions, the effect of the emotional intervention was comprehensively evaluated, and the emotional change trend and sandplay interaction style characteristics of the subjects during multiple interventions were analyzed.

[0039] This implementation method involved a two-week intervention with 30 participants, conducted twice a week, to test and evaluate the method. Subjective questionnaires, EEG signals, sandplay layouts, and interview records were collected to analyze the effectiveness of the intervention method and to analyze the characteristics of sandplay interaction styles.

[0040] Regarding subjective questionnaire assessment, this implementation method evaluated the effect of emotion improvement from two levels: a cross-stage trait-level questionnaire and a state-level questionnaire before and after a single intervention. The results of the trait-level questionnaire are shown in Table 1. Compared with the entire pre-intervention period, after all four interventions, the subjects' FS, STAI-T, PHQ, GAD, SRSS, BAI, and BDI scores all decreased significantly, while CD-RISC and ACS scores increased significantly. This indicates that while the subjects experienced a reduction in negative emotional burdens such as depression, anxiety, and fatigue, their psychological resilience, sleep quality, and attentional control improved, resulting in an overall improvement in mental health. Further comparison of the results after the intervention and during the follow-up period shows that FS, BAI, and BDI scores continued to decrease significantly, while the ACS score further increased. Meanwhile, STAI-T, PHQ, GAD, CD-RISC, and SRSS remained stable, indicating that the intervention effect had good sustainability during the follow-up phase, and some indicators still showed a trend of further improvement. The results of the State Level Questionnaire are shown in Table 2. In the four intervention sessions, after each intervention, the subjects' Valence and Arousal increased from low to medium levels to high levels, STAI-S decreased significantly, PANAS positive emotion scores increased significantly, PANAS negative emotion scores decreased significantly, and NRS pleasantness and NRS relaxation scores also increased significantly. This indicates that a single intervention can promote the release of negative emotions, increase relaxation and positive experience in the subjects, thereby enabling their emotional state to quickly shift towards a more positive and stable direction.

[0041] Table 1: Mean scores of the trait level questionnaire scale Table 2: Mean scores of the State Level Questionnaire Scale Regarding physiological indicators, this implementation method first extracts features from the preprocessed EEG and fNIRS signals, and then analyzes the intervention effect from both EEG and fNIRS perspectives. For the preprocessed EEG signals, this implementation method first performs a 1-45Hz bandpass filter, dividing the signal into time windows of 4 seconds with a 50% overlap. Noise segments are then removed based on amplitude and standard deviation thresholds, and features in the time domain, frequency domain, time-frequency domain, and entropy are calculated. Specifically, a Fast Fourier Transform is performed on the filtered time-domain EEG signal to obtain the corresponding frequency domain spectral representation, which can be expressed as: in: Represents discrete-time EEG signals. N The number of sampling points. The complex spectral representation of the corresponding frequency component is then used; the power spectral density is calculated based on the frequency domain spectrum, and the spectral power is statistically analyzed within a preset frequency band to extract the frequency domain features of the corresponding frequency band.

[0042] The analysis of EEG characteristics included changes before and after the overall intervention, changes before and after a single intervention, and dynamic changes during the intervention process. Firstly, in the comparison of overall EEG changes before and after the intervention, resting-state results showed that after the intervention, the subjects' alpha band differential entropy, relative alpha band power, and alpha / β ratio significantly increased, while alpha band Hjorth complexity, γ band Hjorth mobility, total Hjorth mobility, γ band HFD, total HFD, total sample entropy, γ differential entropy, and relative γ power significantly decreased. Simultaneously, the frontal asymmetry in the delta, high-β, and γ bands significantly decreased (manifested as stronger AF7 than AF8), indicating a more stable resting state after the intervention, a relative increase in low-frequency components, and a decrease in high-frequency components and overall variability. Further observation of the resting-state change trends before each intervention showed that as the intervention progressed, low-β band Hjorth complexity gradually increased, while sample entropy, total Hjorth mobility, and total HFD gradually decreased, suggesting that subjects would exhibit lower expected stress and more stable neurodynamics in later intervention sessions. Secondly, in the EEG characteristics before and after a single intervention, the differential entropy, absolute α power, and α / β ratio all significantly increased after each intervention, while the Hjorth mobility and HFD of the α band significantly decreased. This indicates that a single intervention can increase the proportion of the α component and reduce the volatility of α band activity, demonstrating an immediate relaxation and mood improvement effect. Figure 4 As shown, in the dynamic analysis of the EEG intervention process, the task-state EEG, statistically analyzed in 15-second time windows, showed that the relative α band power of the three channels AF7, AF8, and M2 gradually increased over time, while the θ / α ratio and α band Hjorth complexity gradually decreased. At the same time, the frontal asymmetry of the β and γ bands gradually decreased, indicating that as the intervention progressed, the subjects' emotional fluctuations decreased, task engagement increased, and the α band activity pattern tended to be more regular and stable.

[0043] The OD value, after converting the fNIRS intensity signal into optical density change and removing head motion artifacts, is then converted into hemoglobin concentration change according to the modified Beer-Lambert law, and bandpass filtered at 0.01~0.1Hz to obtain the changes in oxyhemoglobin, deoxyhemoglobin, and total hemoglobin concentrations over time. For example... Figure 5 As shown, during the dynamic process of the intervention, the deoxyhemoglobin HbR concentration in channels 4 and 8 gradually decreased relative to the baseline, indicating a gradual increase in local cerebral blood flow. This reflects the continuous enhancement of the subjects' emotional engagement and immersion over time, demonstrating the superiority of this intervention method.

[0044] In terms of sandplay interaction style feature analysis, this implementation method uses multimodal data to group and identify subjects and perform differential analysis to reveal the interaction style characteristics of different subjects in multiple sandplay interventions. Specifically, firstly, the subjects' interviews and narrative texts are encoded into narrative semantic features using the Qwen embedding model; spatial distribution and object operation descriptors are extracted from the sandplay scene layout as layout features; further, EEG features reflecting neural activity states are extracted as brain activity features; and questionnaire scores are used as emotional state features. Subsequently, low-dimensional representation learning is performed on each modality feature, and the low-dimensional features of each modality are concatenated and fused to form a multimodal joint feature representation of the subjects. Based on this, the K-Means clustering algorithm is used to classify the categories. In this embodiment, three types of sandplay interaction styles were identified: theme-driven, process-driven, and experience-driven, with 7 subjects in the theme-driven category, 12 subjects in the process-driven category, and 11 subjects in the experience-driven category. Theme-driven participants' narratives are typically highly abstract and symbolic, demonstrating the ability to organize and integrate multiple miniature models around a higher-level theme. Their emotional expressions are often calm, focused, and reflective. These participants exhibit the largest number of objects, the widest spatial expansion, and more frequent key model magnification in their scene layouts, tending to construct sand table scenes through combinations of symbolic elements and emphasis on key elements. Process-driven participants' narratives usually possess clear sequence and causal logic. The scene construction process emphasizes rules, orderliness, and a sense of control, and their emotional expressions are often serious and objective. These participants exhibit the fewest magnification operations in their scene layouts, with a moderate number of objects and spatial range, indicating a preference for expressing narrative content through orderly and coherent scene organization. Experience-driven participants' narratives revolve more around individual real-life experiences, life situations, or specific events, with emotional expressions primarily warm, relaxed, and comfortable. These participants exhibit lower levels of object quantity and overall area occupied, using fewer elements and employing a more compact sand table layout, tending to construct realistic scenes around specific characters or life fragments.

[0045] In summary, the above technical solutions demonstrate that this invention's interactive virtual reality sandplay emotion intervention framework based on wearable BCI combines active virtual reality sandplay interaction, AIGC-driven personalized sandplay object generation, and objective monitoring of multimodal EEG signals. This enhances the sense of participation, immersion, and personalized expression during the intervention process, while also improving the objectivity and continuity of emotion assessment. This method can simultaneously support immediate state assessment of a single intervention and long-term effect assessment of multiple interventions, demonstrating significant application value in emotion regulation, psychological function improvement, and intervention process tracking. Furthermore, the integrated interactive hybrid extended reality and wearable BCI intervention approach offers good ease of use and scalability, exhibiting high application value and potential for promotion in areas such as psychological intervention, remote family intervention, and more.

[0046] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. An emotion intervention method based on wearable BCI and interactive mixed extended reality, characterized in that, Includes the following steps: (1) Construct an interactive virtual reality sand table intervention scene, complete the three-dimensional modeling of the sand table space, configure the sand table and sand tool display area, and preset the interactive operation elements for scene construction; (2) Based on AIGC technology, a scalable 3D sandplay tool library is built to generate personalized sandplay tools on demand; (3) The EEG and fNIRS signals of the subjects were collected synchronously through the wearable BCI device during the entire intervention process, and the head motion inertial signal was collected synchronously as a reference to remove motion artifacts in the EEG and fNIRS signals. (4) Establish a structured emotional intervention paradigm to collect multimodal data, including questionnaire information, EEG and fNIRS signals, sand table scene layout, and interview information, at the corresponding stages before intervention, intervention, post-intervention and follow-up. (5) The effect of the emotional intervention was comprehensively evaluated based on the multimodal data obtained during multiple interventions, and the emotional change trend and sandplay interaction style characteristics of the subjects during multiple interventions were analyzed.

2. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The interactive virtual reality sand table intervention scene in step (1) is a virtual room scene immersively constructed using 3D modeling software. It has a central table and a rectangular sand table inside. The surface of the sand table is covered with a layer of sand, and the inner edge of the sand table is set to blue to simulate the "water boundary" in the physical sand table. Shelves for placing sand objects are set on both sides of the sand table. The constructed scene has a wealth of interactive operation elements, including: moving and rotating the viewpoint in a first-person perspective using the controller; adding or deleting water areas on the surface of the sand table using the brush and eraser tools; selecting sand objects from the shelves and placing them in the sand table, and continuously adjusting the posture and scaling the size of the placed sand objects.

3. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The personalized sand figures in step (2) are automatically generated by the AIGC model. The 3D sand figure library includes at least the following 11 basic categories: people, animals, vegetation, buildings, vehicles, fences and signs, natural objects, fantasy objects, spiritual and mysterious objects, landscape components, and household items. The AIGC model generates prompts according to emotional design constraints, including: using soft and harmonious color schemes with moderate contrast; using clear outlines and reasonable proportions for shapes and appearances, avoiding sharp structures and maintaining consistent scale; maintaining a balanced and stable spatial structure; and maintaining a consistent overall style for visual semantics, avoiding violent or uncomfortable appearances. The generated prompt rules are also used for on-demand sandplay creation during the intervention process to achieve dynamic expansion of the sandplay library and personalized symbolic expression. For sandplay generated by AIGC, it will be screened based on structural integrity, appearance consistency and emotional adaptability, and sandplay with strange structure, visual abnormality or inappropriate emotional expression will be removed to form a three-dimensional sandplay library for intervention.

4. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The wearable BCI device in step (3) is a forehead-attached wireless device. The device includes an EEG acquisition module, an fNIRS acquisition module, and an inertial acquisition module. The EEG acquisition module and the fNIRS acquisition module are integrated in the same attachment structure in a stacked manner to achieve synchronous acquisition of EEG and fNIRS signals in the same area of ​​the forehead. The inertial acquisition module includes an accelerometer and a gyroscope, which are used to acquire head motion inertial signals and provide a reference for subsequent EEG and fNIRS signal motion artifact suppression.

5. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 4, characterized in that: In step (3), the EEG signal is collected from four electrodes, Fpz, AF7, AF8 and M2, according to the international 10-20 system standard, with M1 as the reference electrode and a sampling rate of 250Hz; the fNIRS signal is collected in the bilateral forehead region by eight optical probes, using dual wavelengths of 735nm and 850nm and a sampling rate of 25Hz; during the intervention, the head motion inertial signal of 90.91Hz from the virtual reality head-mounted display and 12.5Hz from the wearable BCI device are collected simultaneously as reference signals to adaptively remove artifacts in the EEG signal.

6. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The emotional intervention paradigm in step (4) includes four stages: pre-intervention, intervention, post-intervention, and follow-up. The intervention stage lasts for two weeks, twice a week, for a total of four intervention sessions. The themes are energy, travel, connection, and birth, respectively. Each intervention session includes theme guidance, 20 minutes of virtual reality sand table construction, and review and reflection. The review and reflection includes guided reflection and semi-structured interviews. The sand table scene layout and interview information will be saved after each intervention. Resting-state EEG and fNIRS signals of the subjects are collected in the pre-intervention stage, the post-intervention stage, and before and after each intervention. Task-state EEG and fNIRS signals of the subjects are continuously collected during the virtual reality sand table construction process. At the same time, the subjects' state questionnaire and trait questionnaire scale are collected. Interview and questionnaire information are collected again in the follow-up stage to evaluate the sustainability of the intervention effect.

7. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The questionnaire information in step (4) includes two types: trait questionnaire scale and state questionnaire scale, and the effect of emotion intervention is comprehensively evaluated based on the scores of each scale. Trait questionnaires were collected during the pre-intervention, post-intervention, and follow-up phases, including: The PHQ and BDI are scales used to assess the degree of depression; the higher the score, the more severe the depressive symptoms. GAD, BAI, and STAI-T are scales used to assess anxiety levels; higher scores indicate more severe anxiety. The CD-RISC scale is used to assess psychological resilience; a higher score indicates that an individual has a higher level of psychological resilience and recovery ability. The FS scale is used to assess the degree of fatigue; a higher score indicates more severe fatigue. The SRSS scale is used to assess the severity of sleep problems; a higher score indicates a more pronounced sleep problem. The ACS scale is used to assess attentional control ability; the higher the score, the stronger the attentional control ability. A status questionnaire was collected before and after each intervention session, including: SAM, a scale score used to assess a subject’s current valence and arousal status; The NRS is a scale that assesses the transient intensity of feelings such as pleasure, happiness, calmness, relaxation, anger, disgust, fear, anxiety, sadness, and dizziness. A higher score indicates a higher intensity of the corresponding emotion or state. PANAS is a scale that assesses the current level of positive and negative emotions. A higher score on the positive emotion subscale indicates stronger positive emotions, and a higher score on the negative emotion subscale indicates stronger negative emotions. The STAI-S scale assesses the level of state anxiety; a higher score indicates a more severe level of current anxiety.

8. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: In step (5), the collected EEG and fNIRS signals need to be processed and EEG features extracted. Specifically: First, the first 10 seconds and the last 10 seconds of each signal are removed. The EEG signal is bandpass filtered from 1 to 45 Hz and divided into time window segments with a length of 4 seconds and an overlap rate of 50%. Noise segments are removed based on amplitude threshold and standard deviation threshold. Then, features including kurtosis, HFD, detrended fluctuation analysis parameters, power of each frequency band and frequency band power ratio, Hjorth complexity and mobility features, sample entropy, spectral entropy, and differential entropy are extracted from the processed EEG signal. In addition, based on the frequency band power of AF7 and AF8 channels, the asymmetric features of the EEG prefrontal cortex are calculated. The fNIRS signal is converted and head motion artifacts are removed to obtain the change in optical density. Then, according to the modified Beer-Lambert law, it is converted into the change in hemoglobin concentration and bandpass filtered at 0.01~0.1Hz to obtain the characteristic signals of the changes in oxyhemoglobin, deoxyhemoglobin and total hemoglobin concentrations over time.

9. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The comprehensive evaluation of the emotional intervention effect in step (5) includes two methods: long-term pre- and post-intervention evaluation and single-intervention pre- and post-intervention evaluation. The long-term pre- and post-intervention evaluation method is used to analyze the cumulative effect of multiple interventions. By comparing the questionnaire information and resting-state EEG characteristic signals at the initial stage of the intervention with those after several interventions, the overall trend of the subject's emotional state is evaluated. The single-intervention pre- and post-intervention evaluation method is used to analyze the changes in the subject's state before and after a single sandplay intervention. By comparing the questionnaire information collected before and after the intervention and the resting-state and task-state EEG characteristic signals, the immediate impact of a single intervention on the subject's emotional level and related physiological state is evaluated.

10. The emotion intervention method based on wearable BCI and interactive mixed extended reality according to claim 1, characterized in that: The analysis of sandplay interaction style features in step (5) is based on multimodal data to group and identify subjects. Specifically: First, the narrative text, sandplay scene layout information, EEG features, and questionnaire information of the subjects during the sandplay intervention process are extracted, and corresponding feature representations are constructed respectively. The narrative text is encoded into narrative semantic features through a text embedding model, the sandplay scene layout information is represented as layout features through scene spatial distribution and object operation descriptors, the EEG features are used to represent the neural activity state of the subjects during the intervention process, and the questionnaire information is represented by scale scores to represent the emotional state features of the subjects. Then, the dimensionality reduction of each modal data feature is performed, and the dimensionality reduction multimodal data features are spliced ​​and fused to form a joint feature representation for representing the sandplay interaction style of the subjects. Finally, based on the joint feature representation, a clustering algorithm is used to classify the subjects to distinguish the sandplay interaction styles of the subjects.

Images