Emotion monitoring system based on sleep HRV features

Abstract

This invention discloses an emotion monitoring system based on sleep HRV characteristics, belonging to the field of intelligent health monitoring technology. The system includes: simultaneously collecting PPG, EEG, and near-infrared light signals during sleep; analyzing HRV sequences and respiratory wave morphology based on PPG signals; achieving sleep microstructure staging through cardiopulmonary coupling analysis; extracting dynamic HRV features based on the staging results and comparing them with personalized stage baselines; detecting and labeling emotional load events; constructing and updating a nighttime emotional event atlas; predicting safe intervention time windows based on the atlas and sleep staging; executing personalized interventions and collecting physiological feedback data to form a closed-loop optimization system. This invention achieves high temporal resolution, event-oriented nighttime emotional dynamic identification and safe intervention, improving the accuracy and reliability of emotion monitoring.

Description

Technical Field

[0001] This invention relates to the field of intelligent health monitoring technology, and in particular to an emotion monitoring system based on sleep HRV characteristics. Background Technology

[0002] With advancements in wearable physiological sensing technology and biosignal processing algorithms, emotion monitoring based on heart rate variability (HRV) has become a research hotspot in neuropsychology and health informatics. Traditional methods correlate emotional states by analyzing the statistical characteristics of overnight or long-term HRV (such as time-domain and frequency-domain indicators). In recent years, further attempts have been made to combine HRV analysis with macroscopic sleep stages (such as NREM and REM) in order to improve the context-specificity of emotion assessment.

[0003] Current technologies still have significant limitations. On the one hand, their analytical granularity is coarse, relying heavily on the average HRV of the entire night or macroscopic stages, failing to capture discrete events of high emotional load lasting several minutes induced by specific neural activities (such as micro-awakening and cyclical alternation patterns) during sleep, resulting in the loss of emotional dynamics. On the other hand, existing solutions lack a safe intervention logic compatible with sleep microstructure. If stimulation is recklessly applied to regulate emotions, it can easily lead to sleep fragmentation, impairing core sleep functions and creating an "intervention-interference" paradox. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an emotion monitoring method based on sleep HRV features to address the problem of how to achieve high temporal resolution, event-oriented, accurate identification and safe intervention of nighttime emotional dynamics without damaging sleep structure.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for emotion monitoring based on sleep HRV characteristics, characterized by comprising the following steps: Simultaneously collect PPG signals, EEG signals, and near-infrared light signals during the user's sleep period; HRV sequence and respiratory wave morphology are extracted from PPG signal. Real-time microstructural staging of sleep is performed by calculating the coupling relationship between HRV sequence and respiratory wave morphology, and the staging results are output. Based on the staging results, dynamic feature values ​​are extracted from the HRV sequence within the corresponding sleep stage, and the dynamic feature values ​​are compared with the preset stage baseline threshold to detect and label discrete emotional load events. The marked emotional load events are parsed to obtain their attributes, and the events are correlated with historical events to construct and update the user's nighttime emotional event map. When an emotional load event is detected or predicted based on the emotional event map, the impact of the potential intervention timing on sleep structure is predicted based on the current sleep stage, and an intervention time window that meets the preset safety conditions is found. The corresponding intervention measures are executed within the intervention time window, and physiological feedback data is collected after the intervention. Event attributes, intervention measures and feedback data are stored together to optimize subsequent event detection and intervention decisions.

[0007] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the specific steps for simultaneously acquiring PPG signals, EEG signals, and near-infrared light signals during the user's sleep period are as follows: Using multiple sensors integrated into the head-mounted device, raw PPG waveforms, multi-channel EEG waveforms of the frontal and temporal lobe regions, and dual-wavelength near-infrared light intensity signals of the prefrontal lobe region are continuously acquired at a sampling rate of no less than 256Hz during the user's sleep. All sensor data streams are timestamped with a microsecond-level time base to generate time-aligned raw multimodal signal data packets; Parallel preprocessing is performed on the raw multimodal signal data packets, specifically including motion artifact filtering and baseline drift correction of PPG signals, power frequency interference and electromyography artifact filtering of EEG signals, and ambient light noise suppression of near-infrared light intensity signals. The preprocessed multimodal signals are encapsulated into data frames with time-series markers in real time and transmitted to the subsequent analysis module.

[0008] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the steps include: parsing the HRV sequence and respiratory wave morphology from the PPG signal, calculating the coupling relationship between the HRV sequence and respiratory wave morphology to perform real-time microstructural staging of sleep, and outputting the staging results. From the preprocessed PPG signal, an improved peak detection algorithm is used to locate the peak point of each pulse wave, and a continuous successive heartbeat interval sequence is calculated. The low-frequency oscillation component synchronized with the respiratory rhythm was extracted from the same PPG signal using a morphological decomposition algorithm and used as a respiratory wave morphology sequence. Within a continuous sliding time window, the coherence coefficient and phase synchronization index of the inter-cardiac beat sequence and the respiratory wave morphology sequence are calculated at different frequency bands to obtain the cardiopulmonary coupling strength value. Based on the preset cardiopulmonary coupling strength threshold and dynamic change pattern, the current sleep state is classified in real time into microstructural stages such as stable sleep, unstable sleep or REM sleep, and a staged label stream with time sequence and confidence level is output.

[0009] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the steps include: extracting dynamic feature values ​​from the HRV sequence within the corresponding sleep stage based on the staging results, and comparing the dynamic feature values ​​with a preset stage baseline threshold to detect and label discrete emotional load events. Based on the real-time input stage labels, the feature extraction strategy corresponding to the sleep stage is invoked to calculate the preset time-domain and nonlinear dynamic feature values ​​from the inter-heartbeat sequence; The calculated feature values ​​are compared with the dynamic baseline threshold range corresponding to the same sleep stage, which is established based on the user's historical data, to identify abnormal deviations in the feature values. When the feature value deviates for more than the preset duration and meets the preset pattern, cross-validation is performed by combining the synchronous high-frequency power of EEG or the decreasing trend of prefrontal cortex blood oxygenation. Cross-validated anomalous periods are marked as the start and end points of an emotional load event, and event metadata containing event identifier, occurrence time, duration, and dominant anomalous features is generated.

[0010] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the steps of parsing the marked emotional load events to obtain their attributes and performing correlation analysis between the events and historical events to construct and update the user's nighttime emotional event map are as follows: For each marked event, its intensity level, physiological characteristic combination pattern, and time distance from the sleep stage transition point are quantitatively calculated to form an event attribute vector; The current event attribute vector is compared with previous events in the same sleep cycle to analyze the event clustering phenomenon and intensity evolution trend. The system performs similarity matching and association retrieval of the current event and sequence pattern with events of the same type under the same sleep stage in the user's historical multi-night database. Based on the matching and association results, a visualized event association network graph is dynamically updated in the user's exclusive spatiotemporal coordinate system, with events as nodes and temporal and physiological similarity as edges.

[0011] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the steps of: when an emotional load event is detected or predicted according to an emotional event map, predicting the impact of potential intervention timing on sleep structure based on the current sleep stage, and finding an intervention time window that meets preset safety conditions, are as follows: Upon receiving real-time event alerts or high-risk period signals predicted based on event graph patterns, users can immediately query their current and predicted future sleep stage status. Based on sleep stage status, the probability of micro-arousals or stage transitions being triggered by applying mild stimulation at different future times is read from and calculated from a pre-established model. Starting from the current event time, search backwards to filter out all potential time points where the physiological effects of the event are still within the effective period and the probability of sleep interruption caused by intervention is lower than the first preset safety threshold. From all potential time points that meet the safety criteria, the time closest to the transition point of the next natural sleep stage is selected and ultimately determined as the execution window for this intervention.

[0012] As a preferred embodiment of the emotion monitoring method based on sleep HRV features described in this invention, the steps include: executing corresponding intervention measures within the intervention time window, collecting physiological feedback data after the intervention, and storing the event attributes, intervention measures, and feedback data in association for optimizing subsequent event detection and intervention decisions. At the start of the defined intervention time window, select the intervention type with the highest matching degree to the event attribute from the personalized intervention library, and control the actuator to output acoustic or light modulation stimulation with set parameters; During a fixed monitoring period after the intervention begins, the recovery trajectory of key physiological indicators such as HRV high-frequency power and EEG spectral entropy is continuously collected. The slope of the changes in physiological indicators and the level of homeostasis recovery during the monitoring period before and after the intervention were calculated and quantified as the implicit effect score of this intervention. The complete event attributes, intervention parameters, execution window, and implicit effect score are associated and stored as a training sample in the feedback database for periodic updates to the event detection threshold and intervention strategy matching rules.

[0013] Secondly, the present invention provides an emotion monitoring system based on sleep HRV characteristics, comprising, The signal acquisition module simultaneously acquires PPG signals, EEG signals, and near-infrared light signals during the user's sleep. The sleep staging module analyzes the HRV sequence and respiratory wave morphology from the PPG signal, performs real-time microstructural staging of sleep by calculating the coupling relationship between the HRV sequence and respiratory wave morphology, and outputs the staging results. The event detection module extracts dynamic feature values ​​from the HRV sequence within the corresponding sleep stage based on the stage results, and compares the dynamic feature values ​​with the preset stage baseline threshold to detect and mark discrete emotional load events. The graph construction module parses the marked emotional load events to obtain their attributes and performs correlation analysis between the events and historical events, thereby constructing and updating the user's nighttime emotional event graph. The window optimization module detects or predicts emotional load events based on the emotional event map, predicts the impact of potential intervention timing on sleep structure based on the current sleep stage, and finds an intervention time window that meets preset safety conditions. The feedback optimization module executes corresponding intervention measures within the intervention time window and collects physiological feedback data after the intervention. It associates and stores event attributes, intervention measures, and feedback data to optimize subsequent event detection and intervention decisions.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the emotion monitoring method based on sleep HRV features as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the emotion monitoring method based on sleep HRV features as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By combining cardiopulmonary coupled sleep microstructure staging with HRV dynamic feature extraction, it achieves accurate capture and labeling of transient, discrete high-emotional-load events during sleep, overcoming the problem of loss of emotional dynamic trajectory caused by traditional all-night or macro-staging averaging analysis; by establishing an emotional event atlas that integrates event attributes and historical correlations, it elevates emotion monitoring from state description to pattern recognition, providing a structured and visualized diagnostic basis for understanding the patterns of individual nocturnal emotional fluctuations; by introducing an intervention window optimization mechanism based on sleep structure integrity prediction, it ensures that any regulatory intervention is prudently implemented under the premise of safety without damaging core sleep functions, fundamentally avoiding the risk of sleep fragmentation that may be caused by the intervention itself; by constructing a closed-loop optimization system driven by implicit physiological feedback, the event detection model and intervention strategy can be continuously adaptively optimized based on objective physiological responses, thereby continuously improving the accuracy of personalized monitoring and regulation in long-term applications. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0018] Figure 1 This is a flowchart of the emotion monitoring method based on sleep HRV features in Example 1. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for emotion monitoring based on sleep HRV features, including the following steps: Simultaneously collect PPG signals, EEG signals, and near-infrared light signals during the user's sleep period; HRV sequence and respiratory wave morphology are extracted from PPG signal. Real-time microstructural staging of sleep is performed by calculating the coupling relationship between HRV sequence and respiratory wave morphology, and the staging results are output. Based on the staging results, dynamic feature values ​​are extracted from the HRV sequence within the corresponding sleep stage, and the dynamic feature values ​​are compared with the preset stage baseline threshold to detect and label discrete emotional load events. The marked emotional load events are parsed to obtain their attributes, and the events are correlated with historical events to construct and update the user's nighttime emotional event map. When an emotional load event is detected or predicted based on the emotional event map, the impact of the potential intervention timing on sleep structure is predicted based on the current sleep stage, and an intervention time window that meets the preset safety conditions is found. The corresponding intervention measures are executed within the intervention time window, and physiological feedback data is collected after the intervention. Event attributes, intervention measures and feedback data are stored together to optimize subsequent event detection and intervention decisions.

[0023] It should be noted that the system synchronously collects the user's PPG, EEG, and near-infrared light signals through a head-mounted device, generating high-quality multimodal data packets after time alignment and filtering / denoising. Based on the PPG signals, the HRV sequence and respiratory wave morphology are analyzed, and the coupling strength between the two is calculated to perform refined microstructural staging of sleep. According to the staging results, the system dynamically extracts HRV features for each stage and compares them with the user's personal historical baseline to detect emotional load events, which are then confirmed through multimodal signal cross-validation. For each confirmed event, the system quantifies its attributes to form a vector and performs correlation analysis with events from the same night and historical events to construct and update a visualized emotional event atlas. When an event or atlas predicts high risk, the system combines the current sleep staging with a vulnerability model to proactively find the optimal time window that satisfies both intervention timeliness and ensures sleep safety. Within the safety window, the system matches and executes personalized interventions, then continuously monitors the recovery trajectory of physiological indicators, generates an implicit effect score, and uses the entire process data as a sample to optimize system parameters. This method integrates the entire process from high-precision physiological monitoring, event-based emotion identification, graph-based pattern discovery to safe closed-loop intervention, improving the accuracy and reliability of nighttime emotional dynamic management without damaging the sleep structure.

[0024] Specifically, the steps for synchronously collecting PPG signals, EEG signals, and near-infrared light signals during the user's sleep period are as follows: Using multiple sensors integrated into the head-mounted device, raw PPG waveforms, multi-channel EEG waveforms of the frontal and temporal lobe regions, and dual-wavelength near-infrared light intensity signals of the prefrontal lobe region are continuously acquired at a sampling rate of no less than 256Hz during the user's sleep. All sensor data streams are timestamped with a microsecond-level time base to generate time-aligned raw multimodal signal data packets; Parallel preprocessing is performed on the raw multimodal signal data packets, specifically including motion artifact filtering and baseline drift correction of PPG signals, power frequency interference and electromyography artifact filtering of EEG signals, and ambient light noise suppression of near-infrared light intensity signals. The preprocessed multimodal signals are encapsulated into data frames with time-series markers in real time and transmitted to the subsequent analysis module.

[0025] It should be noted that, through the PPG, EEG, and near-infrared sensor array integrated into the head-mounted device, the system synchronously acquires raw multimodal physiological signals at a sampling rate of no less than 256Hz, and assigns a uniform microsecond-level timestamp to all data points to achieve precise temporal alignment. Subsequently, the system performs parallel preprocessing on the time-aligned signals, filtering out motion artifacts and baseline drift in the PPG signals, power frequency and physiological artifacts in the EEG signals, and ambient light noise in the near-infrared signals to improve signal quality. Finally, the system encapsulates the cleaned signals into standardized, time-stamped data frames and transmits them stably to downstream analysis modules. This process constructs a complete data link from high-fidelity synchronous acquisition, precise temporal alignment, targeted noise suppression to standardized encapsulation and transmission, providing a high-quality, time-consistent multimodal physiological data foundation for subsequent fusion analysis.

[0026] Specifically, the steps involve parsing the HRV sequence and respiratory wave morphology from the PPG signal, calculating the coupling relationship between the HRV sequence and respiratory wave morphology to perform real-time microstructural staging of sleep, and outputting the staging results. From the preprocessed PPG signal, an improved peak detection algorithm is used to locate the peak point of each pulse wave, and a continuous successive heartbeat interval sequence is calculated. The low-frequency oscillation component synchronized with the respiratory rhythm was extracted from the same PPG signal using a morphological decomposition algorithm and used as a respiratory wave morphology sequence. Within a continuous sliding time window, the coherence coefficient and phase synchronization index of the inter-cardiac beat sequence and the respiratory wave morphology sequence are calculated at different frequency bands to obtain the cardiopulmonary coupling strength value. Based on the preset cardiopulmonary coupling strength threshold and dynamic change pattern, the current sleep state is classified in real time into microstructural stages such as stable sleep, unstable sleep or REM sleep, and a staged label stream with time sequence and confidence level is output.

[0027] It should be noted that the system extracts precise successive heartbeat interval sequences from the preprocessed PPG signal using an improved peak detection algorithm, forming the basis for HRV analysis. Next, by filtering or decomposing the same PPG signal in a specific frequency band, the system separates respiratory wave morphology sequences reflecting respiratory rhythm, achieving synchronous acquisition of heartbeat and respiratory information. Then, within a sliding time window, the system calculates the coherence and phase synchronization index of the two sequences in the physiological frequency band, fusing them to generate a quantified "cardiopulmonary coupling" strength value. Finally, the system compares this real-time calculated coupling strength value with a pre-established sleep microstate pattern library, thereby outputting sleep microstructure staging labels with confidence scores. This innovative process utilizes a single PPG signal to complete the entire process from signal processing to cardiopulmonary coupling calculation and refined sleep staging, providing a non-intrusive, accurate, and autonomic nervous system-aligned temporal framework for subsequent emotion analysis.

[0028] Specifically, based on the staging results, dynamic feature values ​​are extracted from the HRV sequence within the corresponding sleep stage, and these dynamic feature values ​​are compared with a preset stage baseline threshold to detect and label discrete emotional load events. The specific steps are as follows: Based on the real-time input stage labels, the feature extraction strategy corresponding to the sleep stage is invoked to calculate the preset time-domain and nonlinear dynamic feature values ​​from the inter-heartbeat sequence; The calculated feature values ​​are compared with the dynamic baseline threshold range corresponding to the same sleep stage, which is established based on the user's historical data, to identify abnormal deviations in the feature values. When the feature value deviates for more than the preset duration and meets the preset pattern, cross-validation is performed by combining the synchronous high-frequency power of EEG or the decreasing trend of prefrontal cortex blood oxygenation. Cross-validated anomalous periods are marked as the start and end points of an emotional load event, and event metadata containing event identifier, occurrence time, duration, and dominant anomalous features is generated.

[0029] It should be noted that the system dynamically calls the corresponding HRV feature calculation subroutine based on the real-time sleep microstructure stage labels to achieve state-specific feature extraction. Subsequently, the calculated real-time feature values ​​are compared with a dynamic baseline threshold constructed from the user's personal historical data for the same stage to identify individualized physiological deviations. When abnormalities consistent with persistence and specific patterns are identified, the system immediately initiates multimodal cross-validation, comparing synchronized EEG or near-infrared signals to confirm consistency. Finally, the validated abnormal periods are formally marked as emotional load events, and structured metadata containing start and end times, duration, and core abnormal features is generated. This process, through stage-guided feature extraction, personalized baseline comparison, multimodal cross-validation, and standardized event generation, constructs a precise, reliable, and traceable automated detection and labeling system for nocturnal emotional events.

[0030] Specifically, the steps involve parsing the marked emotional load events to obtain their attributes, and then correlating these events with historical events to construct and update the user's nighttime emotional event map. For each marked event, its intensity level, physiological characteristic combination pattern, and time distance from the sleep stage transition point are quantitatively calculated to form an event attribute vector; The current event attribute vector is compared with previous events in the same sleep cycle to analyze the event clustering phenomenon and intensity evolution trend. The system performs similarity matching and association retrieval of the current event and sequence pattern with events of the same type under the same sleep stage in the user's historical multi-night database. Based on the matching and association results, a visualized event association network graph is dynamically updated in the user's exclusive spatiotemporal coordinate system, with events as nodes and temporal and physiological similarity as edges.

[0031] It should be noted that for each tagged emotional load event, the system first quantifies its core features into a structured attribute vector containing intensity level, abnormal feature combination patterns, and temporal distance from the sleep stage boundary. Then, the system performs a sequence comparison of this vector with other event vectors that occurred that same night, analyzing the clustering phenomenon and intensity evolution trend of the events over time to obtain real-time dynamic context. Next, the system uses this vector and the analyzed sequence pattern as query conditions to retrieve similar event records of the same type of sleep stage from the user's historical database, achieving personalized pattern matching across time scales. Finally, the system uses all events as nodes, establishing association edges based on physiological similarity and temporal proximity, dynamically constructing and updating a visualized emotional event network graph. This process, through event quantification, sequence analysis, historical matching, and graph-based presentation, transforms discrete physiological abnormal fragments into a structured knowledge network revealing individual emotional fluctuation patterns and inherent laws.

[0032] Specifically, when an emotional load event is detected or predicted based on an emotional event map, the impact of predicting potential intervention timing on sleep structure based on the current sleep stage is used to find an intervention time window that meets preset safety conditions. The specific steps are as follows: Upon receiving real-time event alerts or high-risk period signals predicted based on event graph patterns, users can immediately query their current and predicted future sleep stage status. Based on sleep stage status, the probability of micro-arousals or stage transitions being triggered by applying mild stimulation at different future times is read from and calculated from a pre-established model. Starting from the current event time, search backwards to filter out all potential time points where the physiological effects of the event are still within the effective period and the probability of sleep interruption caused by intervention is lower than the first preset safety threshold. From all potential time points that meet the safety criteria, the time closest to the transition point of the next natural sleep stage is selected and ultimately determined as the execution window for this intervention.

[0033] It should be noted that the window optimization module is immediately activated when the system detects an emotional event or predicts a high-risk period based on the sleep microstructure map. This module first queries and obtains the current and short-term future sleep microstructure stage prediction sequences, providing a precise physiological background for decision-making. Then, the system calls a pre-trained sleep vulnerability model to calculate the probability that intervention at various future time points may trigger sleep disruption. Next, the system searches backward from the event time, filtering out all candidate time points that are both within the effective period of the event's influence and meet the requirement that the probability of sleep disruption is below a safe threshold, forming a safe time set. Finally, the system selects the time closest to the transition point of the next natural sleep stage from this set, determining it as the optimal intervention window. This process, by integrating stage prediction, risk quantification, and timing optimization, achieves forward-looking and refined intervention decisions while ensuring the safety of sleep structure, minimizing potential interference with sleep continuity.

[0034] Specifically, the steps involve executing corresponding intervention measures within the intervention time window, collecting physiological feedback data after the intervention, and storing the event attributes, intervention measures, and feedback data in association to optimize subsequent event detection and intervention decisions. At the start of the defined intervention time window, select the intervention type with the highest matching degree to the event attribute from the personalized intervention library, and control the actuator to output acoustic or light modulation stimulation with set parameters; During a fixed monitoring period after the intervention begins, the recovery trajectory of key physiological indicators such as HRV high-frequency power and EEG spectral entropy is continuously collected. The slope of the changes in physiological indicators and the level of homeostasis recovery during the monitoring period before and after the intervention were calculated and quantified as the implicit effect score of this intervention. The complete event attributes, intervention parameters, execution window, and implicit effect score are associated and stored as a training sample in the feedback database for periodic updates to the event detection threshold and intervention strategy matching rules.

[0035] It should be noted that when the system reaches the preset intervention window, it first matches and executes the optimal audio-visual modulation stimulation from the personalized intervention library based on the characteristic attributes of the current emotional event. Simultaneously with the intervention, the system immediately initiates a fixed-duration feedback monitoring period, continuously tracking the dynamic recovery trajectory of key indicators such as HRV high-frequency power and EEG spectral entropy. After the monitoring period, the system calculates an implicit effect score independent of subjective reports by analyzing the slope of the trajectory changes and the steady-state recovery level. Finally, the system stores the complete event attributes, intervention parameters, and effect score as training samples, which are periodically used to optimize the event detection threshold and intervention matching strategy, thus forming a data-driven closed loop based on physiological feedback, enabling the overall system performance to continuously evolve.

[0036] This embodiment also provides a mood monitoring system based on sleep HRV characteristics, including: The signal acquisition module simultaneously acquires PPG signals, EEG signals, and near-infrared light signals during the user's sleep. The sleep staging module analyzes the HRV sequence and respiratory wave morphology from the PPG signal, performs real-time microstructural staging of sleep by calculating the coupling relationship between the HRV sequence and respiratory wave morphology, and outputs the staging results. The event detection module extracts dynamic feature values ​​from the HRV sequence within the corresponding sleep stage based on the stage results, and compares the dynamic feature values ​​with the preset stage baseline threshold to detect and mark discrete emotional load events. The graph construction module parses the marked emotional load events to obtain their attributes and performs correlation analysis between the events and historical events, thereby constructing and updating the user's nighttime emotional event graph. The window optimization module detects or predicts emotional load events based on the emotional event map, predicts the impact of potential intervention timing on sleep structure based on the current sleep stage, and finds an intervention time window that meets preset safety conditions. The feedback optimization module executes corresponding intervention measures within the intervention time window and collects physiological feedback data after the intervention. It associates and stores event attributes, intervention measures, and feedback data to optimize subsequent event detection and intervention decisions.

[0037] This embodiment also provides a computer device applicable to the emotion monitoring method based on sleep HRV features, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the emotion monitoring method based on sleep HRV features as proposed in the above embodiment.

[0038] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0039] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the emotion monitoring method based on sleep HRV features as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0040] In summary, this invention achieves precise capture and labeling of transient, discrete high-emotional-load events during sleep by combining cardiopulmonary coupled sleep microstructure staging with HRV dynamic feature extraction, overcoming the problem of lost emotional dynamic trajectories caused by traditional all-night or macro-staging averaging analysis; by establishing an emotional event atlas that integrates event attributes and historical correlations, it elevates emotion monitoring from state description to pattern recognition, providing a structured and visualized diagnostic basis for understanding individual nocturnal emotional fluctuation patterns; by introducing an intervention window optimization mechanism based on sleep structure integrity prediction, it ensures that any regulatory intervention is prudently implemented under the premise of safety without damaging core sleep functions, fundamentally avoiding the risk of sleep fragmentation that may be caused by the intervention itself; and by constructing a closed-loop optimization system driven by implicit physiological feedback, the event detection model and intervention strategy can be continuously adaptively optimized based on objective physiological responses, thereby continuously improving the accuracy of personalized monitoring and regulation in long-term applications.

[0041] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A sleep HRV feature-based emotion monitoring method, characterized in that, The method comprises the following steps: Synchronously collecting PPG signals, electroencephalogram signals and near-infrared light signals during the sleep of a user; Analyzing an HRV sequence and a respiratory waveform from the PPG signals, calculating the coupling relationship between the HRV sequence and the respiratory waveform to perform real-time microstructure staging of the sleep, and outputting the staging results; According to the staging results, extracting dynamic characteristic values from the HRV sequence in the corresponding sleep stage, comparing the dynamic characteristic values with preset stage baseline thresholds, detecting and marking discrete emotional load events, analyzing the emotional load events to obtain their attributes, and associating the events with historical events to construct and update a night emotional event map of the user; When an emotional load event is detected or predicted according to the emotional event map, predicting the influence of a potential intervention time on the sleep structure based on the current sleep stage, finding an intervention time window that meets preset safety conditions, and performing corresponding intervention measures in the intervention time window; Collecting physiological feedback data after the intervention, and associating and storing the event attributes, intervention measures and feedback data for optimizing subsequent event detection and intervention decision-making. The synchronous collection of PPG signals, electroencephalogram signals and near-infrared light signals during the sleep of a user comprises the following steps:

2. The sleep HRV feature based emotion monitoring method of claim 1, wherein: Using multiple sets of sensors integrated in a head-mounted device, continuously acquiring original PPG waveforms, multi-channel electroencephalogram waveforms of the frontal and temporal lobes, and dual-wavelength near-infrared light intensity signals in the frontal lobe region of a user during sleep at a sampling rate of no less than 256 Hz; Applying the same time reference microsecond-level time stamp to all sensor data streams to generate time-aligned original multi-modal signal data packets; Performing parallel preprocessing on the original multi-modal signal data packets, including motion artifact filtering and baseline drift correction for PPG signals, power frequency interference and electromyographic artifact filtering for electroencephalogram signals, and ambient light noise suppression for near-infrared light intensity signals; Real-time packaging of the preprocessed multi-modal signals into data frames with time sequence markers, and transmission to subsequent analysis modules. The analysis of the HRV sequence and the respiratory waveform from the PPG signals, the calculation of the coupling relationship between the HRV sequence and the respiratory waveform to perform real-time microstructure staging of the sleep, and the output of the staging results comprise the following steps:

3. The sleep HRV feature based emotion monitoring method of claim 2, wherein: Using an improved peak detection algorithm to locate each pulse wave peak point from the preprocessed PPG signals, and calculating a continuous successive heartbeat interval sequence; Applying a morphological decomposition algorithm to extract a low-frequency oscillation component synchronized with the respiratory rhythm from the same PPG signal as a respiratory waveform sequence; In a continuous sliding time window, calculating the coherence coefficients and phase synchronization indices of the heartbeat interval sequence and the respiratory waveform sequence in different frequency bands to obtain a cardiopulmonary coupling strength value; According to preset cardiopulmonary coupling strength thresholds and dynamic change patterns, real-time classification of the current sleep state into stable sleep, unstable sleep or rapid eye movement microstructure stages, and output of staging labels with time sequence and confidence. ​ 4. The sleep HRV feature based emotion monitoring method of claim 3, wherein: According to the staging result, dynamic feature values are extracted from the HRV sequence in the corresponding sleep stage, and the dynamic feature values are compared with the preset stage baseline threshold to detect and mark discrete emotional load events, and the specific steps are, According to the real-time input of the staging label, the feature extraction strategy corresponding to the sleep stage is called to calculate the preset time domain and nonlinear dynamic feature values from the heartbeat interval sequence; The calculated feature values are compared with the dynamic baseline threshold range corresponding to the same sleep stage established based on the user's historical data to identify the abnormal deviation of the feature values; When the feature value deviation lasts for more than a preset duration and meets a preset mode, it is cross-verified with the synchronous EEG high-frequency power or frontal lobe oxygen reduction trend; The abnormal period after cross-verification is marked as the starting point and ending point of an emotional load event, and event metadata including event identification, occurrence time, duration and dominant abnormal features are generated.

5. The sleep HRV feature based emotion monitoring method of claim 4, wherein: The marked emotional load events are analyzed to obtain their attributes, and the events are associated with historical events for correlation analysis, thereby constructing and updating the user's night emotional event graph, and the specific steps are, For each marked event, the intensity level, physiological feature combination mode, and time distance from the sleep stage transition point are quantitatively calculated to form an event attribute vector; The current event attribute vector is compared with the previous events in the same sleep cycle to analyze the event clustering phenomenon and intensity evolution trend; The current event and sequence mode are similarity matched and associated with events in the same type of sleep stage in the user's historical multi-night database; According to the matching and association results, an event association network graph is dynamically updated in a user-specific spatio-temporal coordinate system, with events as nodes and time sequence and physiological similarity as edges.

6. The sleep HRV feature based emotion monitoring method of claim 5, wherein: When an emotional load event is detected or predicted according to the emotional event graph, the influence of potential intervention timing on sleep structure is predicted based on the current sleep stage, and a safe intervention time window is found, and the specific steps are, Upon receiving the real-time detected event alarm or the high-risk period signal predicted based on the event graph pattern, the current and predicted future sleep staging state is immediately queried; According to the sleep staging state, the probability of triggering micro-awakening or stage transition by applying a slight stimulus at different future times is read and calculated from the pre-established model; Starting from the current event time, all potential time points where the physiological impact of the event is still within the effective period and the sleep interruption probability caused by intervention is lower than a first preset safety threshold are searched and selected; From all potential time points that meet the safety conditions, the time closest to the next natural sleep stage transition point is selected as the execution window of this intervention.

7. The sleep HRV feature based emotion monitoring method of claim 6, wherein: In the intervention time window, the corresponding intervention measures are executed, and physiological feedback data is collected after the intervention, and the event attributes, intervention measures and feedback data are associated and stored for optimizing subsequent event detection and intervention decision-making, and the specific steps are, At the start of the defined intervention time window, select the intervention type with the highest matching degree to the event attribute from the personalized intervention library, and control the actuator to output acoustic or light modulation stimulation with set parameters; During a fixed monitoring period after the intervention begins, the recovery trajectory of key physiological indicators such as HRV high-frequency power and EEG spectral entropy is continuously collected. The slope of the changes in physiological indicators and the level of homeostasis recovery during the monitoring period before and after the intervention were calculated and quantified as the implicit effect score of this intervention. The complete event attributes, intervention parameters, execution window, and implicit effect score are associated and stored as a training sample in the feedback database for periodic updates to the event detection threshold and intervention strategy matching rules.

8. An emotion monitoring system based on sleep HRV features, based on the emotion monitoring method based on sleep HRV features of any one of claims 1-7, characterized in that: include, The signal acquisition module simultaneously acquires PPG signals, EEG signals, and near-infrared light signals during the user's sleep. The sleep staging module analyzes the HRV sequence and respiratory wave morphology from the PPG signal, performs real-time microstructural staging of sleep by calculating the coupling relationship between the HRV sequence and respiratory wave morphology, and outputs the staging results. The event detection module extracts dynamic feature values ​​from the HRV sequence within the corresponding sleep stage based on the stage results, and compares the dynamic feature values ​​with the preset stage baseline threshold to detect and mark discrete emotional load events. The graph construction module parses the marked emotional load events to obtain their attributes and performs correlation analysis between the events and historical events, thereby constructing and updating the user's nighttime emotional event graph. The window optimization module detects or predicts emotional load events based on the emotional event map, predicts the impact of potential intervention timing on sleep structure based on the current sleep stage, and finds an intervention time window that meets preset safety conditions. The feedback optimization module executes corresponding intervention measures within the intervention time window and collects physiological feedback data after the intervention. It associates and stores event attributes, intervention measures, and feedback data to optimize subsequent event detection and intervention decisions. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: When the processor executes the computer program, it implements the steps of the emotion monitoring method based on sleep HRV features as described in any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: When the computer program is executed by the processor, it implements the steps of the emotion monitoring method based on sleep HRV features as described in any one of claims 1 to 7.

Images