A method and system for EEG-based auditory reward assessment using wearable devices
By performing auditory reward tasks on wearable devices, sampling point-level feedback event labels are generated, EEG signals are collected synchronously and valid trials are screened, ERP features are extracted, and a three-axis reward profile is generated. This solves the accuracy and stability problems of EEG assessment in portable scenarios and realizes multi-dimensional reward assessment.
Patent Information
- Application Number
- CN202611179461.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-10
AI Technical Summary
In existing technologies, visual reward tasks are not suitable for portable application scenarios, lack feedback probability and event locking structures, and wearable EEG with few channels is susceptible to artifacts and insufficient trials, making it difficult to unify output reward sensitivity, feedback learning efficiency, and social reward preference.
By using an auditory reward task based on wearable devices, sampling point-level feedback event labels are generated, EEG signals and behavioral response data are collected simultaneously, valid trials are screened, reward-related ERP features are extracted, and a three-axis reward profile is generated.
In wearable scenarios, it improves the accuracy and stability of EEG auditory reward assessment, and can correspond to specific feedback events and behavioral records to generate multi-dimensional reward assessment indicators.
Smart Images

Figure CN123069015A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of EEG auditory reward assessment technology, and in particular to an EEG auditory reward assessment method and system based on a wearable device. Background Technology
[0002] Reward processing capacity reflects an individual's neural responses to positive feedback, negative feedback, learning gains, and social affirmation. RewP can be used to reflect feedback processing related to reward prediction error and reward uncertainty. Acute stress reduces reward-related neural activity, indicating the need to incorporate state factors into the assessment. Different positive, negative, and combined reinforcing feedback can induce different FRN performances, supporting the differentiation of feedback types in this embodiment. Reward memory and depression-related states affect positive reward potentials; therefore, simply reporting the average waveform is insufficient to form a stable profile. Decreased positive affect may affect reward learning by reducing information encoding in positive reward potentials.
[0003] Feedback-related negative potentials (FRNs) remain an important indicator of feedback processing and learning. Recent meta-analyses of FRNs suggest that feedback processing is influenced by cognitive tasks, motor tasks, and feedback attributes. Research on reinforcement learning under aversive outcome conditions shows that the relationship between positive reward potentials and feedback learning is affected by outcome valence and task context. Therefore, current approaches often only provide general results on "reward processing," lacking specific procedures for feedback event locking, effective trial selection, and ERP comparison under different feedback conditions.
[0004] However, the relevant technologies have the following technical problems: 1) Existing visual reward tasks require screen gaze and visual response coordination, which is not suitable for all portable, ear-worn, or head-mounted application scenarios.
[0005] 2) Existing studies mostly report RewP or FRN separately, lacking an engineered process that unifies reward sensitivity, feedback learning efficiency, and social reward preference into an evaluation profile.
[0006] 3) Existing EEG music, emotional feedback, or ordinary auditory feedback solutions lack clear feedback probabilities, feedback types, and event locking structures, and cannot be directly used for reward feedback ERP quantification.
[0007] 4) Wearable EEG with fewer channels is more susceptible to artifacts, wearing status, and insufficient trials. Summary of the Invention
[0008] To address the aforementioned technical problems, the present invention aims to provide a method and system for EEG auditory reward assessment based on wearable devices, which can improve the accuracy of EEG auditory reward assessment results in wearable scenarios with fewer channels.
[0009] The first technical solution adopted in this invention is: a brainwave auditory reward assessment method based on a wearable device, comprising the following steps: Based on wearable devices, auditory reward tasks are performed on the target human body. At the moment the feedback sound begins, sampling point-level feedback event tags are generated, and the target human body's EEG signals and behavioral response data are collected simultaneously. Based on the feedback event tags, EEG signals are extracted and EEG segments are locked. After screening effective auditory reward task feedback trials, reward-related ERP features are extracted. By combining reward-related ERP features and behavioral response data, multi-dimensional reward evaluation indicators are calculated to generate a three-axis reward profile.
[0010] Furthermore, the step of performing an auditory reward task on the target human body based on a wearable device, generating a sampling point-level feedback event label at the start of the feedback sound, and simultaneously collecting the target human body's EEG signals and behavioral response data, specifically includes: Based on wearable devices, auditory reward tasks are performed on the target human body. Each task is configured with corresponding feedback valence, feedback probability and sound material parameters. The auditory reward tasks include auditory probability reward tasks, auditory reward learning tasks and auditory social reward preference tasks. Based on an auditory reward task, an EEG sampling counter is read when each feedback sound begins to play, generating a feedback event label that includes the task type, feedback category, trial number, and sampling point index. The subject's behavioral selection results, reaction time, and task stage data are recorded simultaneously to generate the target human's EEG signal and behavioral response data.
[0011] Furthermore, the step of extracting EEG signals based on feedback event labels, locking EEG segments, and extracting reward-related ERP features after screening effective auditory reward task feedback trials specifically includes: Using the sampling point index in the feedback event label as time zero, the EEG signal is segmented to obtain the feedback-locked EEG segment; The feedback-locked EEG segments were subjected to pre-feedback baseline correction to obtain baseline-corrected EEG segments. The baseline-corrected EEG segments were sequentially subjected to artifact detection, reaction time detection, label integrity detection, and condition quantity detection to obtain a set of effective feedback trials for the auditory reward task. The effective feedback trial set of the auditory reward task is grouped into trial groups, and the RewP features and FRN amplitude features corresponding to each group are extracted to construct reward-related ERP features.
[0012] Furthermore, the auditory probability reward task, whose RewP feature is the difference in positive reward potential, specifically includes: Based on the effective feedback trial set of the auditory reward task, and combined with the feedback valence, it is divided into positive feedback group and non-positive feedback group; Based on the positive feedback group and the non-positive feedback group, the average amplitude of the EEG segments within the preset time window was calculated and the difference was processed to obtain the reward positive potential difference.
[0013] Furthermore, the auditory reward learning task includes early FRN features and late FRN features in its FRN amplitude characteristics, specifically including: Based on the effective feedback trial set of the auditory reward task, the trial sequence is divided into early learning stage and late learning stage. The feedback-related negative wave amplitudes corresponding to the early and late learning stages are extracted to obtain the early FRN features and late FRN features.
[0014] Furthermore, the auditory social reward preference task's RewP features include social reward RewP features and non-social reward RewP features, specifically including: Based on the effective feedback trial set of the auditory reward task, and combined with the feedback attributes, it is divided into social reward feedback group and non-social reward feedback group; For the social reward feedback group and the non-social reward feedback group, the corresponding RewP amplitude and standard deviation are calculated respectively to obtain the social reward RewP feature and the non-social reward RewP feature.
[0015] Furthermore, the step of calculating multi-dimensional reward evaluation indicators by combining reward-related ERP features and behavioral response data to generate a three-axis reward profile specifically includes: The behavioral selection bias parameters of the target human body are calculated based on the difference in positive reward potential and behavioral response data, and then normalized and weighted to obtain the reward sensitivity index. Based on the difference between early and late FRN features, and combined with behavioral response data, the learning rate parameter of the target human body is obtained, and then normalized and weighted to obtain the feedback learning efficiency index. Based on the difference between social reward RewP features and non-social reward RewP features, the social reward preference index is obtained by combining and standardizing the features. Based on the reward sensitivity index, feedback learning efficiency index, and social reward preference index, a three-axis reward profile is generated.
[0016] Furthermore, the step of generating a three-axis reward profile based on the reward sensitivity index, feedback learning efficiency index, and social reward preference index specifically includes: A three-axis coordinate system framework is constructed by using the reward sensitivity index, feedback learning efficiency index, and social reward preference index as three coordinate axes. Based on a three-axis coordinate system framework, preset marking attributes are assigned, including the number of valid trials, the proportion of artifacts, and the quality marking of auditory reward task completion. Based on the quality labeling of auditory reward tasks, the dimensions corresponding to unexecuted tasks are marked as missing. The data and quality labels of the three-axis coordinate system framework are integrated to generate the final three-axis reward profile.
[0017] The second technical solution adopted in this invention is: a brainwave-auditory reward assessment system based on a wearable device, comprising: The first module is used to perform auditory reward tasks on the target human body based on wearable devices, generate sampling point-level feedback event tags at the start of the feedback sound, and simultaneously collect the target human body's EEG signals and behavioral response data. The second module is used to extract EEG signals based on feedback event labels and lock EEG segments, and extract reward-related ERP features after screening effective auditory reward task feedback trials. The third module is used to combine reward-related ERP features and behavioral response data to calculate multi-dimensional reward evaluation indicators and generate a three-axis reward profile.
[0018] The beneficial effects of the method and system of this invention are as follows: This invention performs an auditory reward task on a target human body based on a wearable device, generates a sampling point-level feedback event label at the start of the feedback sound, and simultaneously collects the target human body's EEG signals and behavioral response data; further, it extracts EEG signals based on the feedback event labels and locks the EEG segments, extracts reward-related ERP features after screening effective auditory reward task feedback trials, reads EEG sampling points at the start of the feedback sound and generates feedback sampling point labels; it extracts feedback-locked EEG segments according to the feedback sampling point labels; it filters the feedback-locked EEG segments for artifacts, reaction time, label integrity and condition quantity to form effective feedback trials, which is used to handle the synchronization, stability and interpretability issues in wearable multi-channel scenarios; finally, it calculates multi-dimensional reward evaluation indicators by combining reward-related ERP features and behavioral response data, and generates a three-axis reward profile result. By processing task conditions, feedback event labels, effective trial quality, ERP time window, behavioral learning indicators and three-axis profile in sequence, the reward evaluation results in multi-channel wearable scenarios can be mapped to specific feedback events, ERP windows and behavioral records. Attached Figure Description
[0019] Figure 1 This is a flowchart of the steps of an EEG-based auditory reward assessment method based on a wearable device according to the present invention; Figure 2 This is a structural block diagram of an EEG auditory reward assessment system based on a wearable device according to the present invention; Figure 3This is a schematic diagram of the system evaluation process provided in a specific embodiment of the present invention; Figure 4 This is a timing diagram of an auditory probability reward task provided in a specific embodiment of the present invention; Figure 5 This is a schematic diagram of the auditory reward learning task process provided in a specific embodiment of the present invention; Figure 6 This is a schematic diagram of the social reward preference task process provided in a specific embodiment of the present invention; Figure 7 This is a schematic diagram of a three-axis reward image provided in a specific embodiment of the present invention. Detailed Implementation
[0020] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
[0021] First, the technical terms used in the embodiments of the present invention will be explained: 1) EEG: Electroencephalogram (EEG) refers to weak bioelectrical activity collected by electrodes on the scalp or forehead.
[0022] 2) ERP: Event-related potential, refers to the time-locked neural response extracted from EEG with the stimulus, feedback or response event as the zero point of time.
[0023] 3) Event label: refers to the data structure recorded simultaneously with each stimulus or feedback event, which includes at least the event type, task conditions, trial number, and EEG sampling point index.
[0024] 4) Sampling point level locking: refers to mapping the starting point of auditory events to specific EEG sampling points using an EEG sampling counter as a unified time base.
[0025] 5) Valid test: refers to the test that passes the checks for reaction time, artifacts, electrode contact quality and label integrity.
[0026] 6) Wearable device: refers to a head-mounted, ear-mounted or similar portable device that can simultaneously output auditory stimulation and collect brainwaves; this embodiment does not limit specific commercial models.
[0027] 7) RewP: Reward Positive Potential. In this embodiment, it is used to characterize the difference in ERP induced by positive feedback relative to that induced by non-positive feedback.
[0028] 8) FRN: Feedback-related negative wave. In this embodiment, it is used to characterize negative feedback or feedback expectation violation related ERP changes.
[0029] 9) MRS: Reward Sensitivity Index, which is formed by the difference between reward feedback ERP and behavioral choice bias.
[0030] 10) FLE: Feedback Learning Efficiency Index, which is formed by changes in ERP related to early and late feedback and behavioral learning trends.
[0031] 11) SRPI: Social Reward Preference Index, used to characterize the ERP response advantage of social reward feedback relative to non-social reward feedback.
[0032] In addition, it should be noted that existing methods for wearable EEG reward assessment typically have four types of problems: First, visual reward tasks are not suitable for all auditory scenarios and introduce visual processing components; second, if the start time of the feedback sound is not tied to the EEG sampling point, the RewP and FRN time windows are prone to shift; third, the number of trials for positive, non-positive, neutral, and social / non-social feedback may be unbalanced, and directly comparing ERP amplitudes will produce bias; fourth, if social rewards are mixed with general rewards, it is difficult to distinguish between overall reward sensitivity, feedback learning efficiency, and social feedback preference.
[0033] Based on this, the embodiments of the present invention use an auditory feedback task and maintain the time lock required for ERP analysis through feedback event labels; distinguish different reward processing modes through MRS, FLE and SRPI three-axis indicators; specify task conditions, event labels, trial screening and indicator formulas; and incorporate reaction time, peak-to-peak value, label integrity and number of conditions into the effective trial screening to avoid relying solely on average waveforms.
[0034] Reference Figure 1 This invention provides a brainwave-auditory reward assessment method based on a wearable device, the method comprising the following steps: S100: Based on wearable devices, perform auditory reward tasks on the target human body, generate sampling point-level feedback event tags at the start of the feedback sound, and simultaneously collect the target human body's EEG signals and behavioral response data. Specifically, based on wearable devices, auditory reward tasks are performed on the target human body. Each task is configured with corresponding feedback valence, feedback probability, and sound material parameters. The auditory reward tasks include auditory probability reward tasks, auditory reward learning tasks, and auditory social reward preference tasks. Based on the auditory reward tasks, an EEG sampling counter is read when each feedback sound begins to play, generating a feedback event label containing task type, feedback category, trial number, and sampling point index. The subject's behavioral selection results, reaction time, and task stage data are recorded simultaneously to generate the target human body's EEG signals and behavioral response data.
[0035] In this embodiment, the task paradigm must be determined first, because the ERP needs to know the corresponding prompt, selection behavior, feedback valence, and task stage for each feedback. The complete reward profile executes the auditory probability reward task, auditory reward learning task, and auditory social reward preference task in sequence by default; if only a single dimension needs to be evaluated, only the corresponding task can be executed, and the unexecuted dimension is marked as missing in the three-axis profile.
[0036] The task configuration table records the prompt type, feedback probability, feedback valence, material number, and stage division. Subsequent feedback event labels directly reference these fields. Therefore, for the same feedback-locked EEG, it not only knows when the feedback occurred, but also whether the feedback belongs to a probabilistic reward, reward learning, or social reward preference task.
[0037] S200: Extract EEG signals based on feedback event labels and lock EEG segments based on feedback; after screening effective auditory reward task feedback trials, extract reward-related ERP features. Specifically, using the sampling point index in the feedback event label as time zero, the EEG signal is segmented to obtain feedback-locked EEG segments; the feedback-locked EEG segments undergo pre-feedback baseline correction to obtain baseline-corrected EEG segments; the baseline-corrected EEG segments are then sequentially subjected to artifact checks, reaction time checks, label integrity checks, and condition quantity checks to obtain a set of valid feedback trials for the auditory reward task; the set of valid feedback trials for the auditory reward task is then grouped, and the corresponding RewP features and FRN amplitude features are extracted for each group to construct reward-related ERP features.
[0038] For the auditory probability reward task, the RewP feature is the reward positive potential difference. First, based on the effective feedback trial set of the auditory reward task, the feedback valence is divided into positive feedback group and non-positive feedback group. Then, based on the positive feedback group and the non-positive feedback group, the average amplitude of the EEG segment within the preset time window is calculated and the difference is processed to obtain the reward positive potential difference.
[0039] In this embodiment, the auditory probability reward task is as follows: Figure 4 As shown, the system presents two distinguishable cue sounds, each corresponding to a different probability of positive feedback. After the subject selects one of the cue sounds, the system plays a positive, non-positive, or neutral feedback sound according to a preset probability. The subject does not need to know the probability allocation beforehand; the system determines their sensitivity to reward feedback by locking the ERP based on their selection behavior and feedback. This task generates the data required for ΔRewP, behavioral selection bias (RB), and MRS.
[0040] For auditory reward learning tasks, the FRN amplitude features include early FRN features and late FRN features. First, based on the effective feedback trial set of the auditory reward task, the trial time sequence is divided into early learning stage and late learning stage. Then, the feedback-related negative wave amplitude corresponding to the early learning stage and late learning stage are extracted respectively to obtain early FRN features and late FRN features.
[0041] In this embodiment, the auditory reward learning task is as follows: Figure 5 As shown, the system presents multiple sound options, each corresponding to a different positive feedback probability; the subject gradually learns the optimal option through multiple trials. The system divides the trials into an early learning stage and a late learning stage, compares the changes in FRN or RewP during the learning process, and fits the upward trend of the optimal option selection ratio. This task forms... , Learning rate k and FLE.
[0042] For the auditory social reward preference task, its RewP features include social reward RewP features and non-social reward RewP features. First, based on the effective feedback trial set of the auditory reward task, the feedback is divided into social reward feedback group and non-social reward feedback group according to the feedback attributes. Then, for the social reward feedback group and the non-social reward feedback group, the corresponding RewP amplitude and corresponding standard deviation are calculated respectively to obtain the social reward RewP features and non-social reward RewP features.
[0043] In this embodiment, the auditory social reward preference task is as follows: Figure 6 As shown, the system presents social reward feedback, non-social reward feedback, and neutral feedback. Social reward feedback can be encouraging speech or sounds with social evaluative meaning, while non-social reward feedback can be non-verbal reward sounds; the three types of feedback are matched as closely as possible in terms of loudness, duration, presentation probability, and acoustic complexity. This task is formed... , , And SRPI.
[0044] Feedback event locking is synchronized with task execution. Whenever a feedback sound begins playing, the system immediately reads the EEG sampling counter and writes the sampling point into the feedback event tag. This is because RewP and FRN are ERP components that appear shortly after feedback; if the feedback start point is off-target, the measurement window (250-350ms, etc.) will be misaligned.
[0045] The sampling point at the start of feedback is then used as a common reference point for pre-feedback baseline correction and post-feedback ERP window measurement. The system first uses the pre-feedback baseline window to eliminate potential drift before the start of the trial, and then measures RewP or FRN in the pre-set window after feedback. Since behavioral selection occurs before feedback, the selection record is saved together with the feedback label to link the behavior and EEG response of the same trial.
[0046] Feedback event labels should at least record the trial number, task type, feedback type, feedback valence, sound condition, cue tone category, subject selection, reaction time, feedback start sampling point, task stage, and material number. The feedback start sampling point is used for EEG segmentation; the feedback type is used for grouping positive, non-positive, neutral, social reward, and non-social reward conditions; the task stage is used for early / late learning comparison; and the material number is used to record the specific sound source.
[0047] The system simultaneously saves continuous EEG data, feedback event label tables, behavioral response records, audio recordings, and conversation quality records. These records will collectively determine which feedback trials can be included in the ERP comparison in the next step.
[0048] The feedback event label table determines the segmentation starting point and conditional grouping; behavioral response records determine selection bias and learning rate; audio material records determine the social / non-social feedback source; and conversation quality records determine whether to reduce the confidence level of a certain axis. Subsequent screening and index calculations simultaneously read these records to avoid creating a profile based solely on EEG waveforms or behavioral results.
[0049] Feedback segmentation and trial screening must be completed before the indicator calculation, because All three methods—FLE and SRPI—depend on the inter-conditional ERP differences. The system uses the feedback start sampling point as time zero, preferentially extracts EEG segments from 200ms before feedback to 800ms after feedback, and uses the pre-feedback baseline window for baseline correction.
[0050] The filtered feedback trials are regrouped according to feedback type, task stage, and material category. Probabilistic reward tasks require positive and non-positive / neutral feedback groups, reward learning tasks require early and late feedback groups, and social reward preference tasks require social and non-social reward groups. Only when these groupings meet the minimum number of valid trials are the corresponding MRS, FLE, or SRPI output as complete indicators.
[0051] Each feedback trial undergoes artifact checking, reaction time checking, label integrity checking, and condition quantity checking in sequence. Artifact checking includes peak-to-peak threshold, short-duration abrupt changes, and electrode contact quality; reaction time checking excludes trials that are too fast, timed out, or unresponsive; label integrity checking confirms the presence of the task type, feedback type, feedback start sampling point, and sound conditions; condition quantity checking confirms that there are sufficient valid trials for positive, non-positive, neutral, social reward, and non-social reward conditions. Trials that fail the screening are not included in MRS, FLE, or SRPI calculations, but the reason for rejection is entered into the quality record.
[0052] The selected trials were grouped by feedback type and task stage, forming positive feedback ERPs, non-positive / neutral feedback ERPs, social reward ERPs, non-social reward ERPs, and early / late learning stage ERPs. (Reward positive potential difference quantity) for: ; in, ERP is locked for positive feedback conditions. For feedback conditions that are non-positive or neutral, the ERP is locked, and the optimal measurement window for RewP is 250-350ms. If the number of valid trials for the two types of feedback is unbalanced, the system uses equal-quantity random sampling, conditional weighted averaging, or quality labeling to handle the situation.
[0053] S300 combines reward-related ERP features and behavioral response data to calculate multi-dimensional reward evaluation indicators and generate a three-axis reward profile.
[0054] S310. Based on the difference in reward positive potential and behavioral response data, the behavioral selection bias parameters of the target human body are calculated and normalized and weighted to obtain the reward sensitivity index. In this embodiment, MRS is a reward sensitivity index, used to simultaneously represent neural reward response and behavioral selection bias. MRS is needed because... Reflecting the difference in ERP induced by positive feedback relative to non-positive feedback, RB reflects whether subjects tend to choose options with a high reward probability.
[0055] RB is: ; in, The number of times the high-reward probability option is selected. To select the total number of times effectively.
[0056] MRS is: ; S320. Based on the difference between early and late FRN features, the learning rate parameter of the target human body is obtained by fitting the behavioral response data and normalizing and weighting the data to obtain the feedback learning efficiency index. In this embodiment, FLE is a feedback learning efficiency index, used to represent the subject's ability to adjust choices using feedback during multiple trial-and-error processes. FLE is needed because changes in the early and late stages of FRN reflect changes in feedback processing throughout the learning process, and the learning rate k reflects the learning trend of explicit selection behavior. The system first calculates the orientation correction FRN amplitude in the early and late stages. and Refit the optimal option selection ratio The slope as a function of trials: FLE is: ; S330. Based on the difference between social reward RewP features and non-social reward RewP features, the social reward preference index is obtained by combining and standardizing the features. In this embodiment, SRPI stands for Social Reward Preference Index, used to represent the ERP advantage of social reward feedback relative to non-social reward feedback. SRPI is needed because, even with similar overall reward sensitivity, the relative responses to social feedback may still differ. The system calculates the RewP magnitude under both social and non-social reward conditions. and And calculate the pooled standard deviation: ; SRPI is: ; S340 generates a three-axis reward profile based on reward sensitivity index, feedback learning efficiency index, and social reward preference index.
[0057] Specifically, a three-axis coordinate system framework is constructed by using the reward sensitivity index, feedback learning efficiency index, and social reward preference index as three coordinate axes. Based on the three-axis coordinate system framework, preset labeling attributes are assigned, including the number of effective trials, the proportion of artifacts, and the quality label of auditory reward task completion. According to the quality label of auditory reward task completion, the dimensions corresponding to unexecuted tasks are marked as missing. The data and quality labels of the three-axis coordinate system framework are integrated to generate the final three-axis reward profile.
[0058] In this embodiment, after the three indicators are calculated, the system forms... Figure 7 The three-axis image shown. The MRS axis is derived from the probabilistic reward task. The RB axis is derived from the early-late changes in FRN and the learning rate k in reward-based learning tasks; the SRPI axis is derived from the RewP normalized difference between social and non-social reward conditions. Each axis is accompanied by the source task, ERP time window, number of valid trials, artifact rate, task completion status, and quality markers.
[0059] The generation order of the three-axis profile is consistent with the aforementioned data processing order: task probability and feedback valence determine event labels, event labels determine ERP segments, filtered segments determine RewP / FRN amplitudes, ERP amplitudes and behavioral records determine three indicators, and the three indicators then form the profile coordinates. If a certain axis is missing or of low quality, the profile retains the missing or low-quality state of that axis, without using data from other tasks to replace it.
[0060] If only part of the task is performed, the axis corresponding to the unperformed task is marked as missing and will not be included in the calculation of the complete profile area. The three-axis profile is formed by ERP features and behavioral selection data induced by auditory feedback events. The report retains the key sound conditions, feedback sampling points, ERP windows, and behavioral indicators for each axis, making it easier for users to understand the formation process of each reward dimension.
[0061] Therefore, the three-axis profiling is not simply placing three scores on a graph, but rather compressing feedback events, effective trials, ERP window, behavior selection, and quality markers into three coordinate axes. Each coordinate axis can be traced back to the corresponding task and feedback conditions, making it easier to explain which type of feedback processing the axis value comes from in wearable scenarios with few channels.
[0062] Therefore, as Figure 3 As shown, this embodiment of the invention organizes auditory reward tasks, feedback sampling point labels, feedback locking segments, effective feedback trials, RewP / FRN features, MRS / FLE / SRPI, and a three-axis profile into a continuous data processing chain. The system first configures three types of tasks: auditory probability reward, auditory reward learning, and auditory social reward preference, ensuring each reward dimension has corresponding feedback conditions. When the feedback sound begins playing, the system reads the EEG sampling counter and writes a feedback event label. Then, it extracts EEG segments around the feedback sampling point and filters effective feedback trials. RewP and FRN are extracted within a preset window. MRS, FLE, and SRPI are then calculated by combining behavioral selection bias, learning rate, and social / non-social feedback differences. Finally, a three-axis reward profile is formed. This chain allows each profile axis to correspond to a specific feedback event, ERP time window, behavioral selection record, and quality marker.
[0063] In summary, the embodiments of the present invention have the following technical differences compared to the prior art: First, auditory probability reward, auditory reward learning, and auditory social reward preference tasks are generated. EEG sampling points are read and labeled at the start of the feedback sound. Feedback-locked EEG segments are extracted according to the feedback sampling point labels. These segments are then filtered for artifacts, reaction time, label completeness, and condition quantity to form valid feedback trials. RewP and FRN features are extracted from the valid feedback trials. MRS, FLE, and SRPI are calculated based on RewP / FRN features, behavioral selection bias, learning rate, and social / non-social feedback differences. A three-axis reward profile is generated based on MRS, FLE, SRPI, the number of valid trials, and quality markers, specifically including: 1) Combining auditory reward feedback tasks with sampling point-level locking of feedback events.
[0064] 2) Extraction of RewP, FRN and social reward ERP indicators based on effective trial screening.
[0065] 3) Method for generating three-axis reward profiles using MRS, FLE, and SRPI.
[0066] 4) The hardware serves as a carrier for synchronous stimulation and EEG acquisition, without being limited to specific chips, models, or devices.
[0067] The embodiments of the present invention have the following advantages over the prior art: 1) Compared with solutions that only use visual reward psychology tasks, this embodiment also includes auditory feedback events, feedback start sampling point labels, and feedback locking ERP segments, which are more suitable for ear-worn or head-worn portable use scenarios.
[0068] 2) Compared with solutions that only use ordinary EEG music feedback or conventional ERP averaging analysis, this embodiment also includes feedback condition grouping, effective feedback trial screening, RewP / FRN stable feature extraction, and MRS / FLE / SRPI triaxial image source recording to address synchronization, stability, and interpretability issues in wearable multi-channel scenarios.
[0069] 3) The three-axis profiles display reward sensitivity, learning efficiency, and social reward preference, and form hierarchical fields based on the values of each axis.
[0070] 4) The system processes task conditions, feedback event labels, effective trial quality, ERP time window, behavioral learning indicators and three-axis profiles in sequence, so that the reward evaluation results in the few-channel wearable scenario can be mapped to specific feedback events, ERP windows and behavioral records.
[0071] Finally, the embodiments of the present invention will be explained and illustrated with reference to specific examples: First, an auditory probability reward assessment is conducted. After the subject wears the wearable device, the system performs a basic quality check, including EEG channel noise level, wearing stability, and short resting segment quality. Following the check, the host computer generates an auditory probability reward task containing two types of discriminable cue sounds. These two types of cue sounds correspond to higher and lower positive feedback probabilities, respectively. However, the subject is only informed that they need to complete a sound discrimination or selection task, without being directly informed of the probability allocation.
[0072] Each trial consists of four parts: a prompt tone, the subject's response, a random delay, and auditory feedback. At the start of feedback, the system writes a feedback event label, which includes the trial number, prompt tone category, feedback type, reaction time, feedback sampling point, and task stage. EEG data is segmented into trial segments according to the feedback sampling point, and the pre-feedback baseline is corrected.
[0073] During data processing, the system removes trials with excessively fast responses, excessively long response times, excessively high EEG peak values, or missing labels. The remaining valid trials are grouped into positive and negative feedback groups, and the ΔRewP is calculated within a 250-350ms window. Subsequently, the reward sensitivity level is calculated based on the proportion of high-reward-probability cues selected, and the reward sensitivity level is output.
[0074] If the number of valid trials for a certain feedback condition is insufficient, the system does not output the final MRS, but instead prompts in the report to supplement the data collection or reduce the confidence level. In this embodiment, the system obtains a repeatable feedback-locked ERP index through auditory feedback and sampling point-level locking, reducing the impact of visual tasks and cross-device synchronization on the results. In a preferred implementation, the system also lists the number of valid trials, average reaction time, proportion of trials eliminated, and RewP waveform quality label for each type of feedback in the report. If the difference in the number of trials between positive and negative feedback is too large, the system re-estimates ΔRewP using equal-quantity random sampling or weighted averaging, and retains the original estimate for manual review. This processing is used to reduce the impact of imbalanced condition quantity on ERP amplitude estimation.
[0075] In this embodiment, the task parameters can be adjusted according to the subject's age, hearing status, and reaction ability, such as extending the reaction window, reducing the difficulty of sound discrimination, or increasing the number of practice trials. After the above adjustments, the system still follows the procedures of auditory reward feedback event locking, valid trial screening, feedback locking ERP extraction, and MRS output.
[0076] Furthermore, auditory reward learning assessment was conducted. The system provided three sound options, each with a different probability of positive feedback. A short exercise was given before the task began to ensure that the participants understood the button or touch selection method. In the formal task, the participants made a selection after hearing the option prompt each time, and the system output positive or negative feedback according to the corresponding probability.
[0077] The system divides all valid trials into early and late stages in chronological order, and can also use a sliding window approach to estimate the learning curve. Each feedback event is written to the feedback sampling point to ensure that the comparison of early and late ERP is based on the same time-locking rule.
[0078] During processing, the system calculates the amplitudes of early and late-stage orientation corrections (FRN or RewP) and fits the behavioral learning rate k of the subject's preferred option. If the ERP change and the behavioral learning direction are consistent, the FLE confidence level is high; if they are inconsistent, the report is marked as requiring manual review.
[0079] In this embodiment, the system combines neural and behavioral indicators of feedback learning through the aforementioned processing, avoiding judgment of feedback learning efficiency solely based on accuracy or ERP amplitude. Regarding anomaly handling, if a subject consistently selects the optimal option in the early stages, resulting in a lack of significant upward potential in the learning curve, the system marks this as a high initial performance pattern, primarily referencing changes in FRN or RewP as feedback expectations are established. If a subject's behavioral choices are nearly random, but the ERP indicators show clear feedback differentiation, the system marks the result as a neuro-behavioral inconsistency pattern. This pattern can serve as an alternative implementation to indicate a separation between reward feedback processing and explicit selection behavior.
[0080] In terms of output applications, FLE can not only be presented as a single numerical value, but also output early-to-late ERP difference maps, optimal option selection ratio curves, and effective trial count statistics. This output method helps to illustrate that the report simultaneously displays feedback events, EEG characteristics, and behavioral learning processes, enabling the numerical source of FLE to be mapped from task records to specific data.
[0081] Finally, a social reward preference assessment is conducted, and the system prepares three types of sound: social reward feedback, non-social reward feedback, and neutral feedback. Social reward feedback can be affirmative speech or sounds with social evaluative meaning; non-social reward feedback can be scores, coins, or abstract positive sounds; neutral feedback is sound without reward meaning. The loudness, duration, and frequency of presentation of the three types of feedback are matched as closely as possible.
[0082] In the task, participants completed simple auditory discrimination or reaction tasks. The system output three types of feedback in a pseudo-random order and recorded the sampling point and condition label for each feedback. To avoid the order effect, the number of consecutive feedbacks of the same type was limited.
[0083] The system calculates the RewP magnitudes for social reward and non-social reward conditions separately, and then calculates the SRPI by combining the standard deviations. If the ERP response for social reward feedback is significantly higher than that for non-social reward feedback, the SRPI is positive; otherwise, it indicates a weakening of social reward preference.
[0084] In this embodiment, the system separates social reward processing from general reward sensitivity through the aforementioned processing, enabling the report to distinguish between two different modes: "weak overall reward response" and "weak social feedback-specific response." To reduce ambiguity caused by differences in speech content, social reward speech can be pre-normalized in loudness and controlled in duration, and randomly rotated between different evaluation statements. Non-social reward sounds are also matched for loudness and duration. The system records the specific feedback category and material number in the event label, allowing for subsequent tracking of whether a particular material leads to an abnormal ERP response.
[0085] In the report output, the system interprets SRPI and MRS together: if both MRS and SRPI are low, it indicates a decreased overall reward response; if MRS is normal but SRPI is low, it indicates a weakened relative social feedback response; if SRPI is high but MRS is normal, it indicates an enhanced social reward preference. The above classification is only for status tracking and manual review reference and does not directly provide a medical diagnosis.
[0086] Reference Figure 2 A wearable device-based EEG auditory reward assessment system includes: The first module 201 is used to perform an auditory reward task on the target human body based on a wearable device, generate a sampling point-level feedback event label at the start of the feedback sound, and simultaneously collect the target human body's EEG signals and behavioral response data. The second module 202 is used to extract EEG signals based on feedback event labels and lock EEG segments, and extract reward-related ERP features after screening effective auditory reward task feedback trials. The third module 203 is used to combine reward-related ERP features and behavioral response data to calculate multi-dimensional reward evaluation indicators and generate a three-axis reward profile.
[0087] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0088] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this is not intended to limit the scope of the embodiments of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the embodiments of the present application.
Claims
1. A brainwave-auditory reward assessment method based on wearable devices, characterized in that, Includes the following steps: Based on wearable devices, auditory reward tasks are performed on the target human body. At the moment the feedback sound begins, a sampling point-level feedback event label is generated, and the target human body's EEG signals and behavioral response data are collected simultaneously. Based on the feedback event tags, EEG signals are extracted and EEG segments are locked. After screening effective auditory reward task feedback trials, reward-related ERP features are extracted. By combining reward-related ERP features and behavioral response data, multi-dimensional reward evaluation indicators are calculated to generate a three-axis reward profile.
2. The EEG-auditory reward assessment method based on a wearable device according to claim 1, characterized in that, The step of performing an auditory reward task on a target human body based on a wearable device, generating a sampling point-level feedback event label at the start of the feedback sound, and simultaneously collecting the target human body's electroencephalogram (EEG) signals and behavioral response data, specifically includes: Based on wearable devices, auditory reward tasks are performed on the target human body. Each task is configured with corresponding feedback valence, feedback probability and sound material parameters. The auditory reward tasks include auditory probability reward tasks, auditory reward learning tasks and auditory social reward preference tasks. Based on an auditory reward task, an EEG sampling counter is read when each feedback sound begins to play, generating a feedback event label that includes the task type, feedback category, trial number, and sampling point index. The subject's behavioral selection results, reaction time, and task stage data are recorded simultaneously to generate the target human's EEG signal and behavioral response data.
3. The EEG-auditory reward assessment method based on a wearable device according to claim 2, characterized in that, The step of extracting EEG signals based on feedback event labels, locking EEG segments, and extracting reward-related ERP features after screening effective auditory reward task feedback trials specifically includes: Using the sampling point index in the feedback event label as time zero, the EEG signal is segmented to obtain the feedback-locked EEG segment; The feedback-locked EEG segments were subjected to pre-feedback baseline correction to obtain baseline-corrected EEG segments. The baseline-corrected EEG segments were sequentially subjected to artifact detection, reaction time detection, label integrity detection, and condition quantity detection to obtain a set of effective feedback trials for the auditory reward task. The effective feedback trial set of the auditory reward task is grouped into trial groups, and the RewP features and FRN amplitude features corresponding to each group are extracted to construct reward-related ERP features.
4. The EEG-auditory reward assessment method based on a wearable device according to claim 3, characterized in that, The auditory probability reward task, whose RewP feature is the difference in positive reward potential, specifically includes: Based on the effective feedback trial set of the auditory reward task, and combined with the feedback valence, it is divided into positive feedback group and non-positive feedback group; Based on the positive feedback group and the non-positive feedback group, the average amplitude of the EEG segments within the preset time window was calculated and the difference was processed to obtain the reward positive potential difference.
5. The EEG-auditory reward assessment method based on a wearable device according to claim 4, characterized in that, The auditory reward learning task has FRN amplitude features including early FRN features and late FRN features, specifically including: Based on the effective feedback trial set of the auditory reward task, the trial sequence is divided into early learning stage and late learning stage. The feedback-related negative wave amplitudes corresponding to the early and late learning stages are extracted to obtain the early FRN features and late FRN features.
6. The EEG-auditory reward assessment method based on a wearable device according to claim 5, characterized in that, The auditory social reward preference task includes RewP features comprising social reward RewP features and non-social reward RewP features, specifically including: Based on the effective feedback trial set of the auditory reward task, and combined with the feedback attributes, it is divided into social reward feedback group and non-social reward feedback group; For the social reward feedback group and the non-social reward feedback group, the corresponding RewP amplitude and standard deviation are calculated respectively to obtain the social reward RewP feature and the non-social reward RewP feature.
7. The EEG-auditory reward assessment method based on a wearable device according to claim 6, characterized in that, The step of calculating multi-dimensional reward evaluation indicators by combining reward-related ERP features and behavioral response data to generate a three-axis reward profile specifically includes: The behavioral selection bias parameters of the target human body are calculated based on the difference in positive reward potential and behavioral response data, and then normalized and weighted to obtain the reward sensitivity index. Based on the difference between early and late FRN features, and combined with behavioral response data, the learning rate parameter of the target human body is obtained, and then normalized and weighted to obtain the feedback learning efficiency index. Based on the difference between social reward RewP features and non-social reward RewP features, the social reward preference index is obtained by combining and standardizing the features. Based on the reward sensitivity index, feedback learning efficiency index, and social reward preference index, a three-axis reward profile is generated.
8. The EEG-auditory reward assessment method based on a wearable device according to claim 7, characterized in that, The step of generating a three-axis reward profile based on reward sensitivity index, feedback learning efficiency index, and social reward preference index specifically includes: A three-axis coordinate system framework is constructed by using the reward sensitivity index, feedback learning efficiency index, and social reward preference index as three coordinate axes. Based on a three-axis coordinate system framework, preset marking attributes are assigned, including the number of valid trials, the proportion of artifacts, and the quality marking of auditory reward task completion. Based on the quality labeling of auditory reward tasks, the dimensions corresponding to unexecuted tasks are marked as missing. The data and quality labels of the three-axis coordinate system framework are integrated to generate the final three-axis reward profile.
9. A brainwave-auditory reward assessment system based on a wearable device, characterized in that, Includes the following modules: The first module is used to perform auditory reward tasks on the target human body based on wearable devices, generate sampling point-level feedback event tags at the start of the feedback sound, and simultaneously collect the target human body's EEG signals and behavioral response data. The second module is used to extract EEG signals based on feedback event labels and lock EEG segments, and extract reward-related ERP features after screening effective auditory reward task feedback trials. The third module is used to combine reward-related ERP features and behavioral response data to calculate multi-dimensional reward evaluation indicators and generate a three-axis reward profile.
