A method and system for rapid detection of emotions induced by VR gory scenes based on electroencephalography (EEG)

Abstract

This invention provides a method and system for rapid detection of emotions induced by VR gory scenes based on electroencephalography (EEG), belonging to the field of EEG emotion detection technology. Subjects wear virtual reality devices connected to an 8-channel EEG acquisition device to obtain raw EEG signals with scene time stamps. Time-windowed EEG data is obtained through preprocessing and time window segmentation. Low-frequency energy features are calculated in the 1-10Hz frequency band, and phase coupling features between EEG channels are extracted based on a 2Hz narrowband segmentation. The coupling relationship is distinguished into contralateral long-distance and ipsilateral short-distance coupling based on scalp spatial location, and their intensity features are statistically analyzed. The low-frequency energy features and the long- and short-distance coupling strengths are constructed as composite discriminative features and input into a discriminative model to achieve rapid detection of gory scenes, improving the timeliness, objectivity, and accuracy of emotion recognition.

Description

Technical Field

[0001] This invention relates to the field of EEG emotion detection technology, and in particular to a method and system for rapid detection of emotions induced by VR gory scenes based on EEG. Background Technology

[0002] With the rapid development of virtual reality technology, it has been widely applied in fields such as gaming, medical simulation training, stress intervention, and psychological research. Emotional state plays a crucial role in virtual reality interaction, serving as an important indicator for evaluating the rationality of scene design, the user-friendliness of human-computer interaction, and safety of use. Current technologies primarily rely on subjective scale ratings by participants before, during, or after the experience, such as subjective scoring of scenes based on emotional valence and arousal. This type of method depends on subjective reports and is easily influenced by social expectations, stress responses, and individual differences in expression. Especially in stimulating scenes such as those involving bloodshed or intense fear, it is difficult to provide timely and objective feedback on the participants' true emotional state. Therefore, solutions relying solely on subjective evaluation have significant limitations.

[0003] To overcome the limitations of purely subjective evaluation, current research has begun to incorporate physiological signals such as heart rate, skin conductance, and respiration to assist in emotion recognition, and is attempting to classify emotions using EEG power spectrum features. With the development of wearable devices, wireless brain-computer interfaces, and mobile computing technologies, EEG-based emotion recognition is gradually evolving towards fewer channels, portability, and real-time capabilities. Researchers recognize that emotional processing is not only reflected in changes in energy across certain frequency bands, but also in the synchronicity and functional connectivity patterns between different brain regions. Therefore, features such as phase synchronization, phase coupling, and functional networks are beginning to be used to identify emotions and cognitive states. Simultaneously, for virtual reality environments, the overall trend in emotion monitoring is to integrate EEG acquisition devices with head-mounted displays, transmitting signals wirelessly to the terminal to achieve online monitoring and intelligent control of specific scenarios, such as bloody scenes, to improve experience safety and personalized intervention capabilities.

[0004] Despite the progress made in the aforementioned research, there are still significant shortcomings in the rapid and reliable detection of emotions induced by gory scenes in virtual reality. Most existing methods are designed for general emotion classification and lack a technical approach to establish specific EEG feature patterns for gory scenes, thus limiting the accuracy and specificity of gory scene recognition. Existing solutions often rely on numerous electrode channels and complex connectivity networks, making them difficult to match with the limited-channel portable EEG devices required for practical applications, and they also lack real-time performance. Regarding feature utilization, existing technologies often focus on a single dimension, such as analyzing only the power spectrum or overall connectivity indicators, without systematically revealing and utilizing the combined pattern of energy enhancement, contralateral long-distance coupling enhancement, and ipsilateral short-distance coupling weakening in the low-frequency 1-10 Hz band to identify specific emotions induced by gory scenes in virtual reality. This results in insufficient automatic and rapid detection capabilities for gory scenes in real-world virtual reality applications. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for rapid detection of emotions induced by bloody scenes in VR based on electroencephalography (EEG). This invention solves the problem in existing technologies that it is difficult to identify the emotional state induced by bloody scenes in virtual reality in a timely, objective, and accurate manner.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A rapid detection method for emotions induced by VR gory scenes based on electroencephalography (EEG) includes:

[0008] When the subjects wore virtual reality devices and connected to an 8-channel EEG acquisition device, a bloody scene and a control scene were presented, and the scene time markers were recorded simultaneously to obtain the raw EEG signals with scene time markers.

[0009] The raw EEG signals are preprocessed and divided according to the time window length and time step, combined with the scene time markers, to obtain time window EEG data corresponding to different scene types;

[0010] The time-window EEG data is subjected to frequency domain transformation, and the energy of each EEG channel is calculated in the 1 to 10 Hz frequency band to obtain a low-frequency energy feature set.

[0011] Within the 1 to 10 Hz frequency band, narrow bands are divided with a 2 Hz narrow band width. The phase coupling index between any two EEG channels within each narrow band is calculated to obtain the narrow band EEG channel phase coupling feature set.

[0012] Based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, the phase coupling relationship between any two EEG channels is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship. The phase coupling index corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship is statistically analyzed to obtain the contralateral long-distance coupling strength characteristics and ipsilateral short-distance coupling strength characteristics.

[0013] The low-frequency energy feature set, the long-distance coupling strength feature on the opposite side, and the short-distance coupling strength feature on the same side are used to construct a composite discriminant feature. The composite discriminant feature is then input into the discriminant model to output the bloody scene detection result.

[0014] Preferably, the subject is a human subject, the gory scene is a virtual reality scene containing gory content presented in the virtual reality device, and the control scene is a virtual reality scene without gory content presented in the virtual reality device.

[0015] Preferably, the 8-channel EEG acquisition device is configured with EEG channels in the bilateral frontotemporal region and adjacent regions of the subject's scalp to acquire multi-channel EEG signals from the bilateral frontotemporal region and adjacent regions of the subject's scalp.

[0016] Preferably, the preprocessing of the raw EEG signal includes: bandpass filtering, artifact removal, and reference reconstruction.

[0017] Preferably, the time window length is 2 to 10 seconds, and the time step is 0.5 to 5 seconds.

[0018] Preferably, the steps for calculating the energy of each brainwave channel include:

[0019] The power spectral density of the spectrum corresponding to the EEG data of each time window is integrated in the frequency band from 1 to 10 Hz to obtain the energy of each EEG channel.

[0020] Preferably, the frequency band from 1 to 10 Hz is divided into narrowbands with a width of 2 Hz. The phase coupling index between any two EEG channels within each narrowband is calculated to obtain a narrowband EEG channel phase coupling feature set, including:

[0021] Within the 1Hz to 10Hz frequency band, adjacent narrowband frequency bands are successively divided with a starting frequency of 1Hz and a narrowband width of 2Hz.

[0022] Narrowband bandpass filtering is performed on the EEG data of each EEG channel within each narrowband frequency band to obtain the corresponding narrowband EEG signal;

[0023] The instantaneous phase difference between any two EEG channels is calculated based on the narrowband EEG signal, and the phase coupling index is calculated based on the instantaneous phase difference to obtain the phase coupling feature set of the narrowband EEG channels.

[0024] Preferably, based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, the phase coupling relationship between any two EEG channels is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship. The phase coupling indices corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship are statistically analyzed to obtain the contralateral long-distance coupling strength characteristics and the ipsilateral short-distance coupling strength characteristics, including:

[0025] Based on the scalp coordinate system, the spatial coordinates of each EEG channel on the subject's scalp are obtained, and the scalp channel spacing between any two EEG channels is calculated. The phase coupling relationship between any two EEG channels that span the left and right hemispheres of the subject's scalp and the scalp channel spacing is greater than a preset distance threshold is taken as the contralateral long-distance coupling relationship.

[0026] The phase coupling relationship between any two EEG channels located in the frontotemporal region and adjacent regions on the same side of the subject's scalp and with a scalp channel spacing less than or equal to the preset distance threshold is taken as the ipsilateral short-distance coupling relationship.

[0027] The phase coupling indices belonging to the long-distance coupling relationship on the opposite side and the short-distance coupling relationship on the same side are statistically summed or averaged to obtain the long-distance coupling strength characteristics on the opposite side and the short-distance coupling strength characteristics on the same side.

[0028] Preferably, the low-frequency energy feature set, the contralateral long-range coupling strength feature, and the same-side short-range coupling strength feature are constructed into a composite discriminant feature, and the composite discriminant feature is input into the discriminant model to output the bloody scene detection result, including:

[0029] After normalizing the low-frequency energy feature set, the contralateral long-distance coupling strength feature, and the ipsilateral short-distance coupling strength feature, they are concatenated according to a preset feature order to form a feature vector for characterizing a single EEG data point of the time window.

[0030] The feature vector is used as the composite discriminative feature input to the classification model trained by supervised learning to obtain the probability of bloody scene detection;

[0031] The probability of detecting a gory scene is compared with a preset threshold. When the probability of detecting a gory scene is greater than the preset threshold, the gory scene detection result is output as positive; otherwise, the gory scene detection result is output as negative.

[0032] A rapid detection system for emotions induced by VR gory scenes based on electroencephalography (EEG), comprising:

[0033] The scene presentation and EEG acquisition unit is used to present a bloody scene and a control scene when the subject wears a virtual reality device and is connected to an 8-channel EEG acquisition device, and simultaneously record scene time stamps to obtain raw EEG signals with scene time stamps.

[0034] The EEG preprocessing and time window segmentation unit is used to preprocess the raw EEG signal and segment it according to the time window length and time step combined with the scene time marker to obtain EEG data with time windows corresponding to different scene types.

[0035] The low-frequency energy feature extraction unit is used to perform frequency domain transformation on the time window EEG data, calculate the energy of each EEG channel in the 1 to 10 Hz frequency band, and obtain a low-frequency energy feature set.

[0036] The phase coupling feature calculation unit is used to divide the 1 to 10 Hz frequency band into narrow bands with a narrow band width of 2 Hz, calculate the phase coupling index between any two EEG channels in each narrow band, and obtain the narrow band EEG channel phase coupling feature set.

[0037] The coupling relationship statistical analysis unit is used to divide the phase coupling relationship between any two EEG channels into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship based on the spatial position relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, and to statistically analyze the phase coupling index corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship to obtain the contralateral long-distance coupling strength characteristics and the ipsilateral short-distance coupling strength characteristics.

[0038] The feature fusion and emotion discrimination unit is used to construct a composite discrimination feature from the low-frequency energy feature set, the contralateral long-distance coupling strength feature and the same-side short-distance coupling strength feature, and input the composite discrimination feature into the discrimination model to output the bloody scene detection result.

[0039] The present invention discloses the following technical effects:

[0040] This invention, by simultaneously presenting a bloody scene and a control scene and recording scene time markers under the condition that the subject wears a virtual reality device and is connected to an 8-channel EEG acquisition device, transforms the emotional response induced by the bloody scene in virtual reality into an objective EEG timing signal. This fundamentally overcomes the shortcomings of existing technologies that mainly rely on subjective scale scoring, are easily affected by social expectations and individual expression differences, and are difficult to reflect the true emotional state in a timely and objective manner, thus realizing online objective monitoring of the emotional effect of bloody scenes.

[0041] This invention performs frequency domain transformation on EEG data within a preset frequency band of one to ten Hz, calculates the low-frequency energy characteristics of each EEG channel, and makes a judgment based on the difference in low-frequency energy levels between bloody scenes and control scenes. This addresses the problem in the background technology of lacking a dedicated EEG feature pattern for the specific stimulus of bloody scenes, which leads to limited recognition accuracy and specificity. This invention establishes a low-frequency energy feature pattern that focuses on the emotions induced by bloody scenes, thereby improving the recognition accuracy and specificity of bloody scenes.

[0042] This invention divides the frequency band from 1 to 10 Hz into multiple narrowbands with a 2 Hz narrowband width, and calculates the phase coupling index between any two EEG channels within each narrowband. Furthermore, based on the spatial positional relationship of each EEG channel on the subject's scalp, the phase coupling relationship is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship and statistically analyzed to form contralateral long-distance coupling strength characteristics and ipsilateral short-distance coupling strength characteristics. This makes up for the shortcomings of existing technologies that focus on a single power spectrum or overall connectivity index and do not systematically reveal the combination mode of "low-frequency energy enhancement + contralateral long-distance coupling enhancement + ipsilateral short-distance coupling weakening", so that the specific functional connectivity mode induced by bloody scenes can be quantitatively characterized and utilized.

[0043] This invention employs an 8-channel EEG acquisition device and avoids the problems of high hardware complexity, large computational overhead, and insufficient real-time performance caused by relying on a large number of electrodes and complex whole-brain network calculations in existing methods by grouping and statistically analyzing narrowband phase coupling relationships over long and short distances. While ensuring a low number of channels and portable configuration, it can still extract discriminative functional connectivity features, thus balancing the wearability of the device with the discriminative performance of the algorithm, making it suitable for online applications in embedded or wireless brain-computer interface environments.

[0044] This invention ultimately constructs a composite discriminative feature by combining low-frequency energy feature set, contralateral long-distance coupling strength feature, and ipsilateral short-distance coupling strength feature, and inputs it into the discriminative model to achieve rapid detection of whether the current time window corresponds to a bloody scene in virtual reality. Compared with existing technologies that only utilize a single type of feature and are more sensitive to individual differences and noise, this invention significantly improves the robustness and effectiveness of the detection results through multi-dimensional feature fusion. It can continuously output detection results in a sliding time window manner in virtual reality systems, providing reliable real-time criteria for automatic adjustment of bloody scene content, user safety warnings, and subsequent psychological intervention. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0047] Figure 2 This is a comparison chart of the spectral amplitudes of 8-channel EEG under a bloody scene and a control scene provided in an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this invention provides a method for rapid detection of emotions induced by VR gory scenes based on electroencephalography (EEG), including:

[0051] Step 100: When the subject wears a virtual reality device and is connected to an 8-channel EEG acquisition device, a bloody scene and a control scene are presented, and the scene time markers are recorded simultaneously to obtain the raw EEG signals with scene time markers.

[0052] Step 200: Preprocess the raw EEG signal and divide it according to the time window length and time step combined with the scene time mark to obtain the time window EEG data corresponding to different scene types;

[0053] Step 300: Perform frequency domain transformation on the time window EEG data, calculate the energy of each EEG channel in the 1 to 10 Hz frequency band, and obtain the low-frequency energy feature set;

[0054] Step 400: Divide the frequency band into narrowbands with a 2Hz narrowband width within the 1 to 10Hz frequency band, calculate the phase coupling index between any two EEG channels within each narrowband frequency band, and obtain the narrowband EEG channel phase coupling feature set;

[0055] Step 500: Based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, the phase coupling relationship between any two EEG channels is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship. The phase coupling index corresponding to the contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship is statistically analyzed to obtain the contralateral long-distance coupling strength characteristics and ipsilateral short-distance coupling strength characteristics.

[0056] Step 600: Construct a composite discriminant feature from the low-frequency energy feature set, the long-distance coupling strength feature on the opposite side, and the short-distance coupling strength feature on the same side. Input the composite discriminant feature into the discriminant model and output the bloody scene detection result.

[0057] Specifically, step 100 of this embodiment involves preparing the subject and presenting the scene in a laboratory with controlled ambient lighting. The subject is a healthy adult aged 18 to 40, and is instructed by staff to wear the virtual reality device before the experiment begins. The virtual reality device is a display terminal capable of forming a closed field of view on the head and displaying a three-dimensional immersive image; its function is to present the subject with highly immersive visual content. In this embodiment, the refresh rate of the virtual reality device is preferably set to 72 frames per second to ensure image continuity and reduce visual latency. After the subject enters the virtual reality scene, a bloody scene and a control scene are presented sequentially. The bloody scene is three-dimensional virtual content containing obvious bloodstains and trauma images, while the control scene is a natural environment without bloodstains, such as a typical street scene or indoor scene, used to construct a basic emotional state. To ensure accurate correspondence of time-series data, this embodiment sets a unique scene time stamp for each scene in the virtual reality system to indicate the start and end times of the scene, as well as the scene category label, for subsequent scene type classification of the raw EEG signals.

[0058] This embodiment uses an 8-channel EEG acquisition device to record the brain electrical activity of the subject under different scenarios. The EEG acquisition device is a wearable measurement device that acquires scalp electrical signals by placing electrodes. In this embodiment, eight EEG channels are respectively arranged in the bilateral frontotemporal regions and adjacent areas of the subject's scalp, with four channels on each side. The frontotemporal region refers to the scalp area located at the junction of the frontal and temporal regions, which is closely related to emotional processing and visual stimulus processing. This region was chosen as the signal acquisition location in this embodiment to improve sensitivity to emotional responses evoked by gory images. The center-to-center spacing between the eight acquisition electrodes is preferably set to 20 mm to 30 mm to ensure continuous coverage and appropriate signal spatial resolution. The sampling frequency of the EEG acquisition device is preferably set to 500 Hz to capture brain electrical activity in the low- to mid-frequency range. In this embodiment, the EEG data is transmitted to the recording terminal in real time via wires or wirelessly, and the aforementioned scene time markers are recorded simultaneously, thereby forming raw EEG signals with time correspondence.

[0059] During scene presentation and EEG data acquisition, this embodiment employs a synchronous recording mechanism to ensure temporal consistency between the virtual reality system and the EEG acquisition system. The synchronous recording mechanism is a technical means to ensure that data generated by different devices maintain a corresponding relationship in the time dimension. This embodiment aligns the timelines of the virtual reality device and the EEG acquisition device using a unified trigger signal, ensuring that each scene time marker accurately locates the corresponding sampling point of the raw EEG signal. For example, if a bloody scene lasts 30,000 milliseconds and a control scene lasts 20,000 milliseconds, two continuous raw EEG signals with clear start and end markers are formed at the data recording end. This embodiment continuously monitors the stability of the trigger signal throughout the acquisition process to ensure that the EEG data for each time window can be subsequently processed according to the corresponding scene time markers. The raw EEG signals obtained in this embodiment through the above method have clear scene category labels, start times, and end times, providing an accurate data foundation for subsequent time window division, frequency domain feature extraction, and discrimination model construction.

[0060] Optionally, after acquiring the raw EEG signal in step 100, step 200 of this embodiment preprocesses the raw EEG signal to improve the reliability of subsequent feature extraction and scene discrimination. The preprocessing methods for the raw EEG signal in this embodiment include bandpass filtering, artifact removal, and reference reconstruction. Bandpass filtering refers to retaining only EEG components within a preset frequency range and suppressing excessively low and high frequency noise. For example, in this embodiment, the lower limit frequency of the bandpass filter can be set to 0.5 Hz to 1 Hz, and the upper limit frequency can be set to 40 Hz to 45 Hz to remove DC drift and high-frequency electromyography interference. Artifact removal refers to identifying and weakening or eliminating non-brain-derived interference signals caused by blinking, eye movement, and head movement. In this embodiment, amplitude thresholds and rate of change thresholds can be set to mark and correct portions that significantly exceed normal EEG amplitudes. For example, when a channel experiences abnormal fluctuations exceeding 100 microvolts in a very short time, it is considered an artifact and replaced or interpolated. Reference reconstruction processing refers to the process of recalculating the potential benchmark of each EEG channel according to a preset reference method. In this embodiment, the full-channel average reference method can be adopted, and the average instantaneous potential of the eight EEG channels can be used as the new reference benchmark to reduce the bias caused by a single-point reference and improve the comparability between different channels.

[0061] After preprocessing, this embodiment divides the raw EEG signal based on the time window length and time step, combined with the scene time stamp obtained in step one hundred, to obtain time window EEG data corresponding to different scene types. The time window length refers to the duration of each segment of continuous EEG signal used for analysis on the time axis. In this embodiment, the time window length is preferably set within the range of 2 to 10 seconds; for example, 4 seconds can be selected as the specific time window length. The time step refers to the time interval between the start times of two adjacent time windows. In this embodiment, the time step is preferably set within the range of 0.5 to 5 seconds; for example, 1 second can be selected as the specific time step, allowing overlap between adjacent time windows and thus improving the temporal resolution of the detection. In this embodiment, under a unified time reference, starting from the beginning of the preprocessed EEG signal with scene time markers, continuous EEG segments are sequentially extracted according to the time window length, and the extraction operation is repeated by moving the starting time according to the time step. For each extracted time window EEG data, its start and end times are compared with the scene time markers. Time window EEG data that fall completely within the bloody scene time period are marked as bloody scene type, and time window EEG data that fall completely within the control scene time period are marked as control scene type, thereby forming a time window EEG data set with clear scene category labels, providing input data with clear time sequence and accurate labels for subsequent frequency domain feature extraction steps.

[0062] Further, in step 300 of this embodiment, after obtaining the time-window EEG data with scene category labels, frequency domain transformation is performed on the signals of each EEG channel in each time window to analyze its energy distribution at different frequency components. Frequency domain transformation is a process that converts time-varying EEG signals into an energy representation with frequency as the coordinate, used to reveal the oscillation characteristics of EEG activity in different frequency bands. This embodiment preferably uses a fast spectral analysis method based on piecewise window functions to process the EEG data of each time window to obtain the spectrum corresponding to each EEG channel within that time window. The spectrum refers to the energy density distribution of the EEG signal at various frequency points, which can reflect the energy characteristics of EEG activity in the low-frequency and mid-frequency regions. For example, for an EEG segment with a time window length of 4 seconds and a sampling frequency of 500 Hz, a frequency domain transformation can yield an energy density sequence in the range of 0 Hz to 250 Hz. This embodiment further extracts frequency band data in the range of 1 Hz to 10 Hz from the obtained spectrum in order to focus only on low-frequency EEG activity related to scene-induced emotional responses.

[0063] After obtaining the spectrum of EEG data for the time window, this embodiment calculates the power spectral density of each EEG channel in the 1 Hz to 10 Hz frequency band to obtain the low-frequency energy of that EEG channel. Power spectral density is a quantitative indicator used to describe the average energy intensity of a signal at different frequencies, reflecting the degree of energy concentration of the EEG signal in the target frequency band. This embodiment obtains the low-frequency energy value of each EEG channel by selecting continuous frequency points in the 1 Hz to 10 Hz range in the spectrum and accumulating or numerically integrating the energy densities corresponding to these frequency points. For example, for a certain EEG channel, if it has 18 effective frequency points in the 1 Hz to 10 Hz frequency band, this embodiment accumulates the energy densities of these 18 frequency points sequentially to form the low-frequency energy value of that EEG channel. This embodiment performs the above calculations on 8 EEG channels respectively, and combines the 8 low-frequency energy values ​​corresponding to each time window in a fixed order to form the low-frequency energy feature set of that time window. This feature set is used to characterize the distribution of low-frequency EEG activity intensity in subjects within this time window, providing basic features for subsequent phase coupling feature extraction and discrimination model input.

[0064] Furthermore, after completing the low-frequency energy feature extraction in step 300, in order to further reveal the phase synchronization relationship of EEG signals in the low-frequency range, step 400 of this embodiment performs narrowband division and phase coupling analysis on the time window EEG data within the 1 to 10 Hz frequency band. Specifically, this embodiment first divides the 1 to 10 Hz frequency band into adjacent narrowband frequency bands with a starting frequency of 1 Hz and a narrowband width of 2 Hz. A narrowband frequency band refers to a frequency band with a small frequency span, used for fine analysis of EEG activity within a certain local frequency range. In this embodiment, several adjacent narrowband frequency bands can be obtained, for example, narrowband frequency bands of 1 to 3, 3 to 5, 5 to 7, 7 to 9, and 9 to 10 can be obtained sequentially, with the last narrowband frequency band being truncated according to the upper limit frequency. Through this division method, this embodiment can finely characterize the phase features of EEG within the overall 1 to 10 Hz low-frequency range with a granularity of 2 Hz.

[0065] After completing the narrowband segmentation, this embodiment performs narrowband bandpass filtering on the EEG data of each EEG channel within each narrowband to obtain the corresponding narrowband EEG signal. Narrowband bandpass filtering refers to a filtering method that retains only the EEG components located within a specific target narrowband and suppresses components outside that band, used to separate EEG oscillations within a specific frequency range. In this embodiment, filters are designed for each narrowband to have a relatively flat passband response within the target band and sufficient attenuation outside the passband. For example, the transition bandwidth between the passband edge and the stopband edge should not exceed 0.5 Hz to reduce mutual interference between adjacent frequency bands. By performing narrowband bandpass filtering on the signals of the eight EEG channels in each time window within each narrowband, a narrowband EEG signal set covering all narrowbands and all EEG channels can be obtained, providing high-frequency-resolution input data for subsequent phase coupling calculations.

[0066] In this embodiment, after obtaining each narrowband EEG signal, the instantaneous phase difference between any two EEG channels is calculated based on these narrowband EEG signals, and a phase coupling index is constructed accordingly to obtain a narrowband EEG channel phase coupling feature set. The instantaneous phase difference refers to the difference between the phase values ​​of two narrowband EEG signals at the same time point, used to measure the degree of phase alignment between the two signals at that time point. In this embodiment, the instantaneous phase value at each time sampling point is extracted by performing analytical signal transformation or equivalent processing on the narrowband EEG signals, and then the instantaneous phase difference between any two EEG channels is calculated point-by-point. The stability of these instantaneous phase differences is statistically analyzed over the entire time window to obtain the phase coupling index. The phase coupling index is a quantitative index typically ranging from 0 to 1, used to characterize the degree of phase synchronization between two EEG signals within a certain narrowband frequency band. For example, when the phase coupling index is close to 1, it indicates that the two channels are highly synchronized within that narrowband frequency band; when the index is close to 0, it indicates that there is almost no stable synchronization relationship between the two channels. In this embodiment, phase coupling indices are calculated for each narrowband frequency band and each pair of EEG channels. These indices are then arranged in an orderly manner according to the order of the narrowband frequency bands and the order of the EEG channel combinations, forming a narrowband EEG channel phase coupling feature set that characterizes the multi-channel phase synchronization mode of the current time window in each narrowband frequency band. This provides rich functional connectivity features for subsequent long-distance and short-distance coupling strength statistics and discrimination model input.

[0067] Furthermore, after obtaining the narrowband EEG channel phase coupling feature set in step 400, in order to distinguish functional connectivity features at different spatial scales, step 500 of this embodiment first constructs a scalp coordinate system and calculates the scalp channel spacing based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp. The scalp coordinate system is a coordinate system that uses three-dimensional coordinates to describe the spatial position of each EEG channel, referencing anatomical landmarks on the subject's scalp. Its function is to convert the actual physical position of the electrodes on the scalp into calculable spatial coordinates. In this embodiment, the occipital region, the midline of the forehead, and the two preauricular points can be selected as reference points. The three-dimensional spatial coordinates of each EEG channel are obtained by measurement; for example, the unit of the coordinate components can be millimeters. When the spatial distance between any two EEG channels falls within the range of 20 mm to 100 mm, it is considered a normal channel spacing distribution. A preset distance threshold is used to distinguish between long-distance channel pairs and short-distance channel pairs. In this embodiment, this threshold is set to 60 mm as the basis for dividing contralateral long-distance coupling relationships and ipsilateral short-distance coupling relationships.

[0068] Based on the spatial coordinate acquisition and scalp channel spacing calculation described above, this embodiment classifies the phase coupling relationship between any two EEG channels into contralateral long-distance coupling and ipsilateral short-distance coupling based on the distribution of the eight EEG channels on the subject's scalp. Contralateral long-distance coupling refers to a phase coupling relationship where any two EEG channels are located in the left and right hemispheres of the subject's scalp, respectively, and the scalp channel spacing between them is greater than 60 mm. For example, when the scalp channel spacing between an EEG channel in the left frontotemporal region and an EEG channel in the right frontotemporal region is 80 mm, this embodiment classifies their phase coupling relationship as a contralateral long-distance coupling relationship. Ipsilateral short-distance coupling refers to a phase coupling relationship where any two EEG channels are located in the same side of the subject's frontotemporal region and adjacent areas, and the scalp channel spacing between them is less than or equal to 60 mm. For example, when the scalp channel spacing between two adjacent EEG channels in the same side of the frontotemporal region is 30 mm, this embodiment classifies their phase coupling relationship as an ipsilateral short-distance coupling relationship.

[0069] In one implementation, this embodiment constructs contralateral long-distance coupling strength characteristics and same-side short-distance coupling strength characteristics by statistically summing phase coupling indices belonging to contralateral long-distance coupling relationships and same-side short-distance coupling relationships, respectively. Specifically, for a certain time window and a certain narrowband frequency band, this embodiment filters out all phase coupling indices classified as contralateral long-distance coupling relationships from the feature set obtained in step 400, and accumulates these indices one by one. For example, when there are a total of 10 phase coupling indices belonging to contralateral long-distance coupling relationships within a certain time window, and the value of each index is distributed between 0.3 and 0.9, this embodiment adds these 10 indices sequentially to obtain a feature value used to characterize the long-distance coupling strength of that time window. Similarly, when there are a total of 12 phase coupling indices belonging to same-side short-distance coupling relationships within a certain time window, this embodiment accumulates these 12 indices to obtain a feature value used to characterize the local short-distance synchronization level. The summation method can highlight narrowband frequency bands with higher overall synchronization levels, making the strong coupling window more significantly reflected in the features.

[0070] In another implementation, this embodiment averages the phase coupling indices belonging to contralateral long-distance coupling relationships and contralateral short-distance coupling relationships respectively to construct contralateral long-distance coupling strength features and contralateral short-distance coupling strength features. The averaging method eliminates the influence of differences in the number of phase coupling channels on the feature value range, making it more comparable across different time windows. For example, when there are a total of 8 phase coupling indices belonging to contralateral long-distance coupling relationships within a certain time window, and the sum of their values ​​is 4.8, this embodiment divides 4.8 by 8 to obtain an average value of 0.6 for the contralateral long-distance coupling strength feature. Similarly, when there are a total of 10 phase coupling indices belonging to contralateral short-distance coupling relationships, and the sum of their values ​​is 3.5, this embodiment divides 3.5 by 10 to obtain an average value of 0.35 for the contralateral short-distance coupling strength feature. The averaging method allows for stable comparison of coupling features at different spatial scales across different time windows, improving the reliability of the input features for subsequent discrimination models.

[0071] After completing steps 300 and 500, this embodiment has obtained a low-frequency energy feature set, contralateral long-distance coupling strength feature, and ipsilateral short-distance coupling strength feature to characterize EEG data for a single time window. To eliminate the influence of differences in feature dimensions and numerical ranges on the discrimination results, step 600 of this embodiment first performs normalization processing on the above three types of features. Normalization processing refers to compressing features from different sources into similar numerical ranges through numerical transformation, so that the classification model can process them uniformly. For example, in this embodiment, each energy value in the low-frequency energy feature set can be mapped to the range of 0 to 1 according to its maximum and minimum values, and the contralateral long-distance coupling strength feature and the ipsilateral short-distance coupling strength feature can also be mapped to the range of 0 to 1. After normalization, this embodiment cascades these features according to a preset feature order. Specifically, the low-frequency energy features corresponding to the eight EEG channels can be arranged in a fixed order first, and then a contralateral long-distance coupling strength feature and an ipsilateral short-distance coupling strength feature can be added sequentially to form a feature vector characterizing EEG data for a single time window. A feature vector is a one-dimensional array that combines multiple scalar features in a fixed order within a time window, and is used as input data in a subsequent discriminative model.

[0072] This embodiment uses the aforementioned feature vectors as composite discriminative features input to a classification model trained using supervised learning to obtain the probability of detecting gory scenes. Supervised learning is a method of training a model using samples with correct category labels. In this embodiment, during the model training phase, a large number of time-window EEG data points labeled as gory or control scene types using scene time stamps are selected. Corresponding feature vectors are extracted from each time window, and these feature vectors, along with their respective scene categories, are input into the classification model for parameter optimization. The classification model can be a linear discriminant model, a nonlinear discriminant model, or a multi-layer neural network model. Through multiple rounds of iterative training, the model learns the correspondence between low-frequency energy feature sets, contralateral long-distance coupling strength features, ipsilateral short-distance coupling strength features, and scene categories. In the online detection phase, this embodiment inputs the feature vector of the current time window into the already trained classification model with fixed parameters. The model outputs a scalar value between 0 and 1, which represents the probability of detecting a gory scene, indicating the confidence level that the current time window belongs to a gory scene.

[0073] This embodiment further compares the probability of detecting gore scenes with a preset threshold to output the final gore scene detection result. The preset threshold is a reference value selected in the range of 0 to 1, used to convert the probability output by the model into two discrimination results: positive or negative. In this embodiment, the preset threshold can be set to 0.5 or adjusted in the range of 0.3 to 0.7 according to actual application needs. When the probability of detecting gore scenes corresponding to a certain time window is greater than the preset threshold, this embodiment determines that the EEG data corresponding to that time window is positive for gore scenes and outputs a positive gore scene detection result; when the probability of detecting gore scenes is less than or equal to the preset threshold, this embodiment determines that the EEG data corresponding to that time window is negative for gore scenes and outputs a negative gore scene detection result. By repeatedly performing the above steps for multiple adjacent time windows in a continuous time, this embodiment can perform near real-time tracking of the subject's exposure to gore scenes in a virtual reality environment, providing objective quantitative basis for adjusting the intensity of gore content, psychological risk warning, and subsequent intervention.

[0074] Figure 2 This diagram illustrates the spectral amplitude distribution of EEG signals obtained through an 8-channel EEG acquisition device within the frequency range of 1 to 48 Hz when presenting bloody scenes, achievement cognition scenes, airplane scenes, resting states, and mountainous terrain scenes in a virtual reality device. Each sub-plot in the diagram corresponds to one EEG channel, and the curves of different colors within each sub-plot represent the average spectral variation trends of the five scene types. As can be seen from the diagram, all channels exhibit significant energy distribution in the low-frequency band, and differences are observed between different scenes, providing a visual basis for the frequency domain transformation and energy calculation in step 300.

[0075] like Figure 2 As shown, compared to resting scenes and other non-bloody virtual reality scenes, the amplitude of the 1-10Hz frequency band is significantly enhanced across multiple channels in bloody scenes. This embodiment utilizes this variation characteristic, extracting a low-frequency energy feature set in step 300, and then constructing a composite discriminant feature in subsequent steps 400-600 together with narrowband phase coupling features and long- and short-distance coupling strength features to identify whether the EEG data of the current time window corresponds to a bloody scene. Therefore, Figure 2 This directly demonstrates the low-frequency energy difference basis upon which the technical solution of this invention relies, and verifies the effectiveness of this frequency band in detecting emotional induction in bloody scenes.

[0076] Corresponding to the above method, this embodiment also provides a rapid detection system for emotions induced by VR gore scenes based on electroencephalography (EEG), including:

[0077] The scene presentation and EEG acquisition unit is used to present a bloody scene and a control scene when the subject wears a virtual reality device and is connected to an 8-channel EEG acquisition device, and simultaneously record scene time stamps to obtain raw EEG signals with scene time stamps.

[0078] The EEG preprocessing and time window segmentation unit is used to preprocess the raw EEG signal and segment it according to the time window length and time step combined with the scene time marker to obtain EEG data with time windows corresponding to different scene types.

[0079] The low-frequency energy feature extraction unit is used to perform frequency domain transformation on the time window EEG data, calculate the energy of each EEG channel in the 1 to 10 Hz frequency band, and obtain a low-frequency energy feature set.

[0080] The phase coupling feature calculation unit is used to divide the 1 to 10 Hz frequency band into narrow bands with a narrow band width of 2 Hz, calculate the phase coupling index between any two EEG channels in each narrow band, and obtain the narrow band EEG channel phase coupling feature set.

[0081] The coupling relationship statistical analysis unit is used to divide the phase coupling relationship between any two EEG channels into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship based on the spatial position relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, and to statistically analyze the phase coupling index corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship to obtain the contralateral long-distance coupling strength characteristics and the ipsilateral short-distance coupling strength characteristics.

[0082] The feature fusion and emotion discrimination unit is used to construct a composite discrimination feature from the low-frequency energy feature set, the contralateral long-distance coupling strength feature and the same-side short-distance coupling strength feature, and input the composite discrimination feature into the discrimination model to output the bloody scene detection result.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0084] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, 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 the present invention.

Claims

1. A rapid detection method for emotions induced by VR gory scenes based on electroencephalography (EEG), characterized in that, include: When the subjects wore virtual reality devices and connected to an 8-channel EEG acquisition device, a bloody scene and a control scene were presented, and the scene time markers were recorded simultaneously to obtain the raw EEG signals with scene time markers. The raw EEG signals are preprocessed and divided according to the time window length and time step, combined with the scene time markers, to obtain time window EEG data corresponding to different scene types; The time-window EEG data is subjected to frequency domain transformation, and the energy of each EEG channel is calculated in the 1 to 10 Hz frequency band to obtain a low-frequency energy feature set. Within the 1 to 10 Hz frequency band, narrow bands are divided with a 2 Hz narrow band width. The phase coupling index between any two EEG channels within each narrow band is calculated to obtain the narrow band EEG channel phase coupling feature set. Based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, the phase coupling relationship between any two EEG channels is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship. The phase coupling index corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship is statistically analyzed to obtain the contralateral long-distance coupling strength characteristics and ipsilateral short-distance coupling strength characteristics. The low-frequency energy feature set, the long-distance coupling strength feature on the opposite side, and the short-distance coupling strength feature on the same side are used to construct a composite discriminant feature. The composite discriminant feature is then input into the discriminant model to output the bloody scene detection result.

2. The method for rapid detection of emotions induced by VR gory scenes based on EEG as described in claim 1, characterized in that, The subjects are human subjects, the gory scene is a virtual reality scene containing gory content presented in the virtual reality device, and the control scene is a virtual reality scene without gory content presented in the virtual reality device.

3. The method for rapid detection of emotions induced by VR gory scenes based on EEG as described in claim 1, characterized in that, The 8-channel EEG acquisition device sets up the EEG channels in the bilateral frontotemporal region and adjacent regions of the subject's scalp to acquire multi-channel EEG signals from the bilateral frontotemporal region and adjacent regions of the subject's scalp.

4. The method for rapid detection of emotions induced by VR gory scenes based on EEG as described in claim 1, characterized in that, The preprocessing methods for the raw EEG signals include: bandpass filtering, artifact removal, and reference reconstruction.

5. The method for rapid detection of emotions induced by VR gory scenes based on EEG according to claim 1, characterized in that, The time window length is 2 to 10 seconds, and the time step is 0.5 to 5 seconds.

6. The method for rapid detection of emotions induced by VR gory scenes based on EEG according to claim 1, characterized in that, The steps for calculating the energy of each brainwave channel include: The power spectral density of the spectrum corresponding to the EEG data of each time window is integrated in the frequency band from 1 to 10 Hz to obtain the energy of each EEG channel.

7. The method for rapid detection of emotions induced by VR gory scenes based on EEG as described in claim 1, characterized in that, Within the 1 to 10 Hz frequency band, narrowband frequency bands are divided with a 2 Hz narrowband width. The phase coupling index between any two EEG channels within each narrowband frequency band is calculated to obtain a narrowband EEG channel phase coupling feature set, including: Within the 1Hz to 10Hz frequency band, adjacent narrowband frequency bands are successively divided with a starting frequency of 1Hz and a narrowband width of 2Hz. Narrowband bandpass filtering is performed on the EEG data of each EEG channel within each narrowband frequency band to obtain the corresponding narrowband EEG signal; The instantaneous phase difference between any two EEG channels is calculated based on the narrowband EEG signal, and the phase coupling index is calculated based on the instantaneous phase difference to obtain the phase coupling feature set of the narrowband EEG channels.

8. The method for rapid detection of emotions induced by VR gory scenes based on EEG according to claim 1, characterized in that, Based on the spatial positional relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, the phase coupling relationship between any two EEG channels is divided into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship. The phase coupling indices corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship are statistically analyzed to obtain the contralateral long-distance coupling strength characteristics and the ipsilateral short-distance coupling strength characteristics, including: Based on the scalp coordinate system, the spatial coordinates of each EEG channel on the subject's scalp are obtained, and the scalp channel spacing between any two EEG channels is calculated. The phase coupling relationship between any two EEG channels that span the left and right hemispheres of the subject's scalp and the scalp channel spacing is greater than a preset distance threshold is taken as the contralateral long-distance coupling relationship. The phase coupling relationship between any two EEG channels located in the frontotemporal region and adjacent regions on the same side of the subject's scalp and with a scalp channel spacing less than or equal to the preset distance threshold is taken as the ipsilateral short-distance coupling relationship. The phase coupling indices belonging to the long-distance coupling relationship on the opposite side and the short-distance coupling relationship on the same side are statistically summed or averaged to obtain the long-distance coupling strength characteristics on the opposite side and the short-distance coupling strength characteristics on the same side.

9. The method for rapid detection of emotions induced by VR gory scenes based on EEG as described in claim 1, characterized in that, The low-frequency energy feature set, the contralateral long-range coupling strength feature, and the same-side short-range coupling strength feature are used to construct a composite discriminant feature. This composite discriminant feature is then input into a discriminant model to output the bloody scene detection results, including: After normalizing the low-frequency energy feature set, the contralateral long-distance coupling strength feature, and the ipsilateral short-distance coupling strength feature, they are concatenated according to a preset feature order to form a feature vector for characterizing a single EEG data point of the time window. The feature vector is used as the composite discriminative feature input to the classification model trained by supervised learning to obtain the probability of bloody scene detection; The probability of detecting a gory scene is compared with a preset threshold. When the probability of detecting a gory scene is greater than the preset threshold, the gory scene detection result is output as positive; otherwise, the gory scene detection result is output as negative.

10. A rapid detection system for emotions induced by VR gory scenes based on electroencephalography (EEG), characterized in that, include: The scene presentation and EEG acquisition unit is used to present a bloody scene and a control scene when the subject wears a virtual reality device and is connected to an 8-channel EEG acquisition device, and simultaneously record scene time stamps to obtain raw EEG signals with scene time stamps. The EEG preprocessing and time window segmentation unit is used to preprocess the raw EEG signal and segment it according to the time window length and time step combined with the scene time marker to obtain EEG data with time windows corresponding to different scene types. The low-frequency energy feature extraction unit is used to perform frequency domain transformation on the time window EEG data, calculate the energy of each EEG channel in the 1 to 10 Hz frequency band, and obtain a low-frequency energy feature set. The phase coupling feature calculation unit is used to divide the 1 to 10 Hz frequency band into narrow bands with a narrow band width of 2 Hz, calculate the phase coupling index between any two EEG channels in each narrow band, and obtain the narrow band EEG channel phase coupling feature set. The coupling relationship statistical analysis unit is used to divide the phase coupling relationship between any two EEG channels into contralateral long-distance coupling relationship and ipsilateral short-distance coupling relationship based on the spatial position relationship of each EEG channel of the 8-channel EEG acquisition device on the subject's scalp, and to statistically analyze the phase coupling index corresponding to the contralateral long-distance coupling relationship and the ipsilateral short-distance coupling relationship to obtain the contralateral long-distance coupling strength characteristics and the ipsilateral short-distance coupling strength characteristics. The feature fusion and emotion discrimination unit is used to construct a composite discrimination feature from the low-frequency energy feature set, the contralateral long-distance coupling strength feature and the same-side short-distance coupling strength feature, and input the composite discrimination feature into the discrimination model to output the bloody scene detection result.

Images