Physiological-psychological fusion adaptive sleep regulation method and system

Abstract

The application discloses a physiological-psychological fusion adaptive sleep regulation system and method. The method comprises the following steps: collecting electroencephalogram signals in real time; determining a sleep stage and evaluating a cortical arousal level based on the electroencephalogram signals; during a sleep onset period, executing a physiological degradation first strategy according to the arousal level: outputting physical full stimulation when the arousal level is high, starting cognitive guidance and switching to a cooperative maintenance mode after the low arousal level is stabilized, and rolling back protection; during a light sleep period, executing a subthreshold maintenance strategy: stopping cognitive guidance and only enhancing stimulation for a short time when micro-arousal occurs; during a deep sleep period, executing a slow wave phase synchronization strategy: outputting pulse stimulation synchronized with the phase of slow oscillation; during a rapid eye movement period, executing double silence protection: forcibly setting stimulation to zero and locking after two consecutive windows are confirmed, and releasing after three consecutive windows are exited; during an awakening period, executing a smooth arousal strategy: simulating sunrise light and natural sound gradient output. The application solves the problems of physiological-psychological intervention disconnection, REM period error stimulation and skill migration.

Description

Technical Field

[0001] This invention relates to a physiological-psychological integrated adaptive sleep regulation method, and also to a corresponding adaptive sleep regulation system, belonging to the field of sleep intervention technology. Background Technology

[0002] Sleep disorders, especially chronic insomnia, have complex pathophysiological mechanisms, generally considered to involve multiple dimensions of abnormalities, including cortical hyperarousal, autonomic nervous system dysfunction, and circadian rhythm dysregulation. Epidemiological surveys show that the global prevalence of insomnia among adults is approximately 10%–20%, while in my country, this proportion is as high as 38.2%, making sleep disorders a significant and serious public health problem.

[0003] Current mainstream interventions for chronic insomnia fall into two main categories. The first is multimodal physical neuromodulation techniques, such as transcranial microcurrent stimulation (CES), photo-optical brainwave entrainment, thermal stimulation, and vibration stimulation. These techniques directly regulate neural activity through external physical means, attempting to passively reduce cortical excitability. The second is digital cognitive behavioral therapy (CBT-I), which serves as a first-line treatment for chronic insomnia. It aims to fundamentally improve sleep patterns by correcting patients' misconceptions and unhealthy behaviors regarding sleep.

[0004] However, both of these mainstream approaches have inherent and insurmountable drawbacks. Regarding physical neuromodulation techniques, they can only provide immediate symptomatic intervention, failing to address the cognitive and behavioral root causes of insomnia, thus hindering the production of lasting therapeutic effects. More seriously, the lack of precisely stage-appropriate physical stimulation makes it prone to mis-triggering during specific sleep stages (such as REM sleep), leading to sleep fragmentation and further deteriorating sleep quality. As for digital CBT-I, its effectiveness heavily relies on patients acquiring and internalizing cognitive and behavioral skills during the day. However, insomniacs are often in a state of high arousal at night, during which the prefrontal cortex is significantly suppressed, preventing them from effectively utilizing cognitive skills acquired during the day, thus greatly diminishing the intervention's effectiveness. Furthermore, existing CBT-I programs generally lack effective positive reinforcement mechanisms, resulting in poor user compliance and a high risk of device dependence.

[0005] In summary, the core challenges currently facing sleep regulation technology lie in two aspects: firstly, simple physical interventions only address the symptoms and not the root cause, and are prone to disrupting sleep structure due to inappropriate stimulation; secondly, cognitive behavioral therapy alone is ineffective in the high-arousal state at night. More importantly, existing solutions have failed to organically integrate physiological and psychological interventions, and to switch and coordinate them at the appropriate time and in the appropriate manner based on the user's real-time state. Furthermore, how to establish an adaptive regulation framework that can finely adapt to different sleep stages and possess self-evolutionary capabilities across the diurnal time dimension remains a pressing technical challenge to be solved in this field. Summary of the Invention

[0006] The primary technical problem to be solved by this invention is to provide a physiological-psychological integrated adaptive sleep regulation method.

[0007] Another technical problem to be solved by the present invention is to provide a physiological-psychological integrated adaptive sleep regulation system.

[0008] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: According to a first aspect of the present invention, a physiological-psychological integrated adaptive sleep regulation method is provided, comprising the following steps: Real-time acquisition of users' brainwave signals; The user's current sleep stage is determined based on the EEG signals, which include the wakefulness stage, the sleep onset stage, the light sleep stage, the deep sleep stage, and the REM sleep stage; and the user's cortical arousal level is assessed in real time based on the Beta wave power in the EEG signals. When the user is in the sleep stage, a physiological downgrade-first strategy is executed based on the cortical arousal level: when the cortical arousal level is higher than a preset threshold, a full physical stimulation command is output; when the cortical arousal level drops below the preset threshold and remains below it for a preset time, a cognitive guidance initiation command is output, and the physical stimulation is switched to a coordinated maintenance mode; the coordinated maintenance mode means that the output intensity of acoustic and light stimulation is reduced to below the user's perception threshold, and electrical stimulation is maintained at a low intensity level below the full output during the sleep stage; furthermore, if the cortical arousal level rises above the preset threshold again during the cognitive guidance output process, the cognitive guidance output is immediately paused and the full physical stimulation output is resumed. When the user is in a light sleep period, a preset subthreshold maintenance strategy is executed: cognitive guidance is stopped, and short-term enhanced stimulation is triggered only when micro-awakening is detected; When the user is in deep sleep, a preset slow-wave phase synchronization strategy is executed: pulse stimulation synchronized with the slow oscillation phase of the brain is output, thereby enhancing slow-wave activity; When the user is in the REM (Rapid Eye Movement) phase, a preset dual-silence protection strategy is executed: after two consecutive phase windows confirm the entry into the REM phase, all physical stimulus outputs are forced to zero and locked; after three consecutive phase windows confirm the exit from the REM phase, the lock is released. When the user is in the waking period, a preset smooth wake-up strategy is executed: simulating the gradual change of sunrise light and the gradual output of natural ambient sound.

[0009] Preferably, the adaptive sleep regulation method further includes: When a sleep phase transition is detected, the phase transition gradient protocol is executed; The phase switching gradual protocol includes: the stimulation parameters exiting the previous phase linearly return to zero within 10 seconds, and the stimulation parameters entering the new phase linearly increase to the target baseline value within 5 seconds from zero; and when a switch between the rapid eye movement (REM) phase and the non-REM phase is detected, the dual silencing protection strategy of the REM phase is executed first, and the phase switching gradual protocol is paused.

[0010] Preferably, the adaptive sleep regulation method further includes: Collect the user's cardiovascular signals and / or body movement signals, and detect micro-arousal events based on the cardiovascular signals and / or body movement signals; When a micro-awakening event is detected during light sleep, a short-term enhanced physical stimulus output is triggered, which automatically returns to the subthreshold level after a preset time.

[0011] Preferably, the adaptive sleep regulation method further includes a day-night coordinated dual-closed-loop adaptive iterative step: A nighttime micro-loop is implemented to drive the adjustment of stimulation parameters based on real-time sleep stage results; The morning macro loop is executed, the sleep quality index is calculated based on the overnight sleep data, and the intervention baseline parameters of the nighttime micro loop are dynamically adjusted for that night. A weekly macro-loop is implemented, and based on the sleep data of the whole week, the overall intervention plan for the following week is reset through multiple standardized paths.

[0012] Preferably, the standardized path includes at least: residual sleep onset difficulty path, new early awakening path, insufficient deep sleep path, insufficient compliance path and hierarchical transition path; Among them, the residual difficulty falling asleep path corresponds to the scheme reset of adjusting the stimulation parameters during the sleep onset period, the newly emerging early awakening path corresponds to the scheme reset of adjusting the morning light stimulation time, the insufficient deep sleep path corresponds to the scheme reset of adjusting the intensity of slow wave enhancement stimulation during deep sleep, the insufficient compliance path corresponds to the scheme reset of adjusting the training difficulty and trigger threshold, and the hierarchical transition path corresponds to the scheme reset of reducing the overall stimulation intensity.

[0013] Preferably, the adaptive sleep regulation method further includes positive reinforcement and decoupling steps: When any successful event is detected, dual-channel reinforcement feedback is triggered, including: the physiological reinforcement channel outputs a gentle vibration pulse, and the cognitive reinforcement channel displays positive voice and visual feedback; the successful events include: the nighttime sleep quality index reaching or improving at the physiological level, and the completion of daytime relaxation training and improvement of heart rate variability at the cognitive level. When a successful event is recognized simultaneously at both the physiological and cognitive levels, an enhanced feedback is triggered. When a physiological success event is triggered for three consecutive days, the intensity baseline of the physical stimulus is automatically reduced by 5%, with the cumulative reduction not exceeding 50% of the initial baseline.

[0014] Preferably, the preset threshold is 120% of the user's resting-state EEG Beta wave baseline; when the user's EEG Beta wave relative power is higher than 120% of the baseline value, it is determined to be a high arousal state; when the user's EEG Beta wave relative power is lower than 120% of the baseline value and lasts for 10 seconds, it is determined to be a low arousal stable state.

[0015] Preferably, determining the user's current sleep stage based on the electroencephalogram (EEG) signals specifically includes: The EEG signals are input into a pre-trained CNN-BiLSTM hybrid deep learning model in a 32-second time window; Time-domain and frequency-domain features are extracted using a three-layer convolutional neural network; The time-domain and frequency-domain features are input into a two-layer bidirectional long short-term memory network to extract the temporal context features of sleep stage changes; Based on the temporal context features, a five-category sleep stage result is output, which is one of the following: wakefulness, sleep onset, light sleep, deep sleep, and REM sleep.

[0016] According to a second aspect of the present invention, a physiological-psychological integrated adaptive sleep regulation system is provided, comprising: A multimodal sensing module is used to collect the user's electroencephalogram (EEG) signals in real time; The real-time sleep staging engine, connected to the multimodal perception module, is used to determine the user's current sleep stage based on the EEG signals. The sleep stages include wakefulness, sleep onset, light sleep, deep sleep, and REM sleep. A cortical arousal level detection unit, connected to the multimodal perception module, is used to assess the user's cortical arousal level in real time based on the Beta wave power in the electroencephalogram (EEG) signal. The sleep stage-specific intervention rule base stores the range of physical stimulation parameters corresponding to each sleep stage, the contraindication rules for each stage, and the temporal pairing strategies corresponding to each stage. The temporal pairing strategies include: the physiological degradation priority strategy for the sleep onset stage, the subthreshold maintenance strategy for the light sleep stage, the slow wave enhancement strategy for the deep sleep stage, the dual silence protection strategy for the REM sleep stage, and the smooth arousal strategy for the wakefulness stage. The physiological-psychological temporal coupling intervention module connects the real-time sleep staging engine and the cortical arousal level detection unit. It is used to execute intervention decisions based on the user's sleep stage and cortical arousal level, according to the temporal pairing strategy corresponding to each stage in the sleep stage-specific intervention rule base. The multimodal execution module is used to output physical stimuli and / or cognitive guidance content according to instructions.

[0017] Compared with the prior art, the present invention has the following technical effects: (1) The embodiments of the present invention use a time-coupled intervention module that is activated only during the wakefulness / sleep period to make dynamic decisions based on the level of cortical arousal: when there is high arousal, only physical full stimulation is output to quickly downgrade physiological arousal; after low arousal stabilizes, cognitive guidance is initiated and switched to the collaborative maintenance mode (sound and light are below the perception threshold, and electrical stimulation is low in intensity) to achieve collaborative intervention of "physiological downgrade first → cognitive guidance follow-up" and set a rollback protection mechanism, which effectively solves the core problem that users cannot call cognitive skills in the high arousal state at night.

[0018] (2) The embodiments of the present invention establish a specific intervention rule base covering the sleep onset period, light sleep period, deep sleep period, REM period, and wakefulness period, and independently configure stimulation parameters, taboo rules, and timing pairing strategies for each stage. Furthermore, for the REM period, a dual silent protection is adopted, which combines two consecutive window confirmations to zero-lock with three consecutive window confirmations to exit and unlock, and is combined with a stage switching gradual protocol (10 seconds to zero / 5 seconds to rise to baseline), which fundamentally avoids sleep fragmentation caused by sudden stimulation changes or accidental triggering of the REM period.

[0019] (3) The embodiments of the present invention construct a nested architecture of nighttime micro-closed loop (second-level real-time control), morning macro-closed loop (daily sleep quality index-driven parameter fine-tuning) and weekly macro-closed loop (five standardized path reset schemes), realizing daily iteration and weekly optimization of the intervention scheme, and completely solving the problem of long-term effect decay caused by stimulation tolerance in the prior art.

[0020] (4) The embodiments of the present invention are based on dual-pathway (physiological level: sleep quality index reaches / improves; cognitive level: daytime training is completed and heart rate variability is improved) to identify successful events, trigger dual-channel reinforcement feedback (gentle vibration + App badge and voice), and automatically lower the baseline of stimulation intensity (5% each time, cumulative ≤50%) when the physiological target is met for three consecutive days. At the same time, the training content without equipment is added, realizing the deep coupling of behavior reinforcement and dedependency.

[0021] (5) The present invention uses daytime breathing-stimulation coupling training (carrier frequency = 3 × breathing frequency, increasing by 8% at the end of inhalation and resetting at the end of exhalation) and completely anchors the coupling parameters with the nighttime sleep intervention program. After 7 to 14 days of training, users only need to adjust their breathing to the training frequency to evoke a relaxation experience, successfully transferring relaxation skills to the device-free scenario and breaking through the dependence on hardware. Attached Figure Description

[0022] Figure 1 The overall flowchart of a physiological-psychological integrated adaptive sleep regulation method provided in the first embodiment of the present invention; Figure 2 This is a flowchart of the real-time sleep staging engine in the first embodiment of the present invention; Figure 3 This is a time-series diagram of the positive reinforcement feedback mechanism in the first embodiment of the present invention; Figure 4 This is an overall structural diagram of a physiological-psychological integrated adaptive sleep regulation system provided in the second embodiment of the present invention; Figure 5 This is a schematic diagram of the time-series pairing strategy corresponding to each stage of the specific intervention rule base in the second embodiment of the present invention; Figure 6 This is a schematic diagram of a breathing coupling anchoring and skill generalization training method provided in the third embodiment of the present invention. Detailed Implementation

[0023] The technical content of the present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0024] The core technical concept of this invention lies in constructing a dual-closed-loop nested sleep regulation system that integrates physiological and psychological aspects and coordinates day and night sleep patterns, using real-time sleep staging and cortical arousal level detection as the core feedback signals. Specifically, it establishes a refined intervention rule base specific to each sleep stage (including physiological downgrading during sleep onset, subthreshold maintenance during light sleep, slow-wave phase synchronization during deep sleep, dual silent protection during REM sleep, and smooth awakening during wakefulness). It also introduces a physiological-psychological temporal coupling intervention module that only takes effect during wakefulness / sleep onset (dynamically switching full physical stimulation, cognitive guidance and coordinated maintenance, and rollback protection based on cortical arousal level). This is supplemented by a three-layer nested adaptive decision engine for day and night sleep, positive reinforcement and dedependency coupling, and respiratory anchoring skill transfer. This systematically solves the common industry problems of physiological-psychological separation, neglect of sleep stage intervention contraindications, day-night separation, and inability to transfer skills in existing technologies. The following detailed description of the implementation of this invention is provided in conjunction with specific embodiments.

[0025] First Embodiment like Figure 1 As shown, the first embodiment of the present invention provides a physiological-psychological integrated adaptive sleep regulation method, which specifically includes the following steps: S1: Real-time acquisition of users' multimodal physiological signals.

[0026] When a user wears a device (such as a head-mounted eye mask) and activates sleep mode, the system first uses the multimodal sensing module 1 (such as...) Figure 4 (As shown) Real-time acquisition of the user's physiological signals. In this embodiment, the multimodal perception module 1 includes: an electroencephalogram (EEG) acquisition unit 11, a cardiovascular signal acquisition unit 12, and a behavioral signal acquisition unit 13.

[0027] Specifically, the EEG acquisition unit 11 uses a single-channel frontal dry electrode (Fpz-A1 / posterior ear reference lead) to acquire raw EEG signals at a sampling rate of 250Hz. These signals are then processed by hardware high-pass filtering (0.5Hz), low-pass filtering (35Hz), and digital notch filtering (50 / 60Hz) to obtain clean EEG timing data. Simultaneously, the cardiovascular signal acquisition unit 12 (ear clip-on PPG sensor, 100Hz sampling rate) acquires photoplethysmography (PPG) signals, and the behavioral signal acquisition unit 13 (triaxial accelerometer, 50Hz sampling rate) acquires body movement and position signals.

[0028] The aforementioned multimodal data will be used in subsequent steps for sleep staging (step S2), cortical arousal level assessment (step S3), and micro-arousal detection (step S5). All signal processing is completed at the device edge, and the raw data is not transmitted back to the cloud to protect user privacy.

[0029] S2: Real-time sleep stage determination.

[0030] After the user's multimodal physiological signals are collected based on the above step S1, the collected EEG time-series data is continuously sent to the real-time sleep staging engine 2 (e.g., Figure 4 (As shown). In this embodiment, the real-time sleep staging engine 2 internally deploys a CNN-BiLSTM hybrid deep learning model that has undergone knowledge distillation and compression. The model size is no more than 5MB and can complete a single inference on the device with a latency of ≤100ms.

[0031] like Figure 2 As shown, the real-time sleep staging engine 2 uses a 32-second time window to first extract temporal and frequency domain features from EEG data through a three-layer convolutional neural network (CNN). These features are then input into a two-layer bidirectional long short-term memory network (BiLSTM) to capture the temporal context of sleep stage changes. Finally, it outputs a five-category sleep staging result: wakefulness, sleep onset, light sleep, deep sleep, or REM sleep. This real-time staging result is the core basis for subsequent stage-specific intervention decisions.

[0032] S3: Real-time assessment of the user's cortical arousal level.

[0033] In parallel with sleep staging, cortical arousal level detection unit 3 (such as...) Figure 4 As shown, time-frequency analysis is performed on the same EEG time-series data to calculate the relative power of the Beta band (13-30Hz) in real time, thereby determining whether the user is in a high-arousal state or a low-arousal stable state.

[0034] Specifically, upon first use, the system collects an EEG sample during a resting state (2 minutes of relaxation with eyes closed) to establish a personal Beta wave power baseline. The percentage of real-time Beta wave power relative to the baseline value (Beta wave percentage) is then used as a quantitative indicator of cortical arousal level. When this Beta wave percentage reaches or exceeds 120%, it is considered a high arousal state; when the Beta wave percentage drops below 120% and remains below for 10 seconds, it is considered a low arousal stable state. In this embodiment, the cortical arousal level indicator is specifically used for temporal coupling decisions during sleep onset (described in detail below).

[0035] S4: Physiological-psychological temporal coupling intervention during the sleep onset period.

[0036] When the real-time sleep staging engine 2 determines that the current stage is the sleep onset stage (N1), it enters the time-series coupling intervention process. At this time, the physiological-psychological time-series coupling intervention module 6 (such as...) Figure 4 (As shown) The output content is dynamically determined based on the percentage of Beta waves output by the cortical arousal level detection unit 3.

[0037] If the percentage of Beta waves is ≥120% (i.e., a high arousal state), it indicates that the user's cortex is overactive and cannot effectively receive cognitive guidance. Therefore, the physiological-psychological temporal coupling intervention module 6 only outputs the full physical stimulus command, instructing the multimodal execution module 7 (such as...) Figure 4 (As shown) The system outputs a combination of sound, light, and electrical stimulation. For example, in one embodiment of the invention, this combined stimulation is as follows: a bone conduction speaker emits binaural beats or isochronous sounds, with the frequency sweeping downwards from the user's current alpha peak (12Hz→8Hz, rate 0.1Hz / s), superimposed with a 30dB powder noise substrate; a red LED array outputs pulsed red light (620~660nm) at a frequency of 0.5Hz, with a brightness not exceeding 50 lux; and a CES microcurrent module outputs a current of 100~200μA at 0.5Hz. It is understood that the purpose of this combined stimulation is to rapidly reduce the physiological arousal level, creating conditions for subsequent cognitive intervention. At this time, the system does not output any cognitive behavioral therapy (CBT-I) content.

[0038] As physical stimulation takes effect, the user's cortical arousal level gradually decreases. When the percentage of Beta waves drops below 120% and this low arousal state is maintained stably for 10 seconds, the physiological-psychological temporal coupling intervention module 6 outputs a cognitive guidance initiation command, while simultaneously switching the physical stimulation to a co-maintenance mode. In the co-maintenance mode, the output intensity of acoustic and light stimulation is reduced below the user's perception threshold (e.g., the pink noise floor is reduced to below 25 dB, and the light stimulation brightness is reduced to below 10 lux), while electrical stimulation is maintained at a low intensity level of 50–100 μA. At this time, the multimodal execution module 7 begins to output cognitive guidance content, such as guiding the user to perform diaphragmatic breathing, progressive muscle relaxation, or body scan audio through voice prompts, helping the user transition from physiological calm to psychological relaxation, thereby smoothly entering sleep.

[0039] It should be noted that this embodiment of the invention also includes a rollback protection mechanism. Specifically, during the cognitive guidance output process, if the cortical arousal level detection unit 3 detects that the percentage of Beta waves has risen back to ≥120%, it indicates that the user may have re-entered a hyperarousal state due to anxiety or interference. At this time, the physiological-psychological temporal coupling intervention module 6 will immediately pause the cognitive guidance output and resume the full physical stimulation output, prioritizing physiological downgrading (the percentage of Beta waves drops to the preset value of 120%). This mechanism effectively avoids ineffective intervention when the user fails to receive cognitive guidance.

[0040] Table 1 Comparison of cognitive guidance and physiological stimulation in the N1 stage

[0041] All the instructions output by the aforementioned sleep intervention (physical stimulus parameters, cognitive content identifiers, and switching sequence) are transmitted in real time to the multimodal execution module 7, which drives the corresponding hardware to execute. At the same time, these execution records (such as high arousal duration, whether cognitive guidance was successful, and whether rollback was triggered) are stored locally as input data for subsequent day-night dual-loop iterations (see step S10 for details).

[0042] S5: From the sleep onset period to the light sleep period, implement the subthreshold maintenance strategy.

[0043] When the real-time sleep staging engine 2 detects that the user's sleep stage has transitioned from the sleep onset stage (N1) to the light sleep stage (N2), the system determines that the user has successfully established sleep. At this point, any intervention that may cause interference needs to be avoided. Therefore, the physiological-psychological temporal coupling intervention module 6 immediately stops outputting cognitive guidance and switches the intervention strategy to the pre-set light sleep stage subthreshold maintenance strategy in the rule base.

[0044] Specifically, under this subthreshold maintenance strategy, the multimodal execution module 7 only outputs physical stimuli below the user's perception threshold (e.g., powder noise 25-30 dB, CES microcurrent 50-100 μA, graphene heating plate maintaining a constant temperature of 33-34 °C), without outputting any cognitive content, thereby maximizing the protection of the established sleep structure.

[0045] Furthermore, during the execution of the subthreshold maintenance strategy in light sleep, the system also needs to handle potential micro-arousals. For this purpose, the micro-arousal detection unit 4 (such as...) Figure 4 (As shown) Based on the cardiovascular signal (PPG) and body motion signal (accelerometer) collected by the multimodal sensing module 1, a gliding detection is performed with a step length of 1 second. A micro-awakening event is determined when any two of the following conditions are met: ① The duration of alpha waves (8-13Hz) on brainwaves is ≥3 seconds (this information can be obtained from the real-time sleep staging engine 2); ② Heart rate increases by ≥10% from the baseline value of the current sleep stage (calculated in real time based on PPG signal); ③ Body acceleration amplitude ≥ 0.1g (based on a triaxial accelerometer).

[0046] Once a micro-awakening event is detected, the system temporarily exceeds the subthreshold baseline, triggering a short-term boost in physical stimulation output. For example, acoustic stimulation is increased by 3 dB from the original baseline for 5 seconds, then automatically returns to the subthreshold level. The purpose of this short-term boost is to suppress the further spread of micro-awakening and prevent it from evolving into full wakefulness. Simultaneously, each micro-awakening event is recorded and included in the next day's sleep report for subsequent morning macro-closed-loop parameter adjustments.

[0047] S6: Implement a slow-wave phase-synchronized stimulation strategy during deep sleep.

[0048] When the real-time sleep staging engine 2 determines that the user has entered deep sleep (N3), the system executes a slow-wave phase synchronization strategy. Specifically, deep sleep is characterized by slow brain oscillations (0.5–1 Hz, i.e., delta waves), which are closely related to memory consolidation and metabolic waste clearance. To enhance slow-wave activity, the system detects the peak phase of slow oscillations in the user's brain in real time. When the peak is predicted to be approaching (usually 50–100 ms before the peak), the physiological-psychological temporal coupling intervention module 6 outputs a pulse stimulation command synchronized with the slow oscillation phase.

[0049] The multimodal execution module 7 receives the pulse stimulation command output by the physiological-psychological temporal coupling intervention module 6, emitting a pulse tone of 0.75–0.8 Hz (volume ≤35 dB) and a CES microcurrent (150–300 μA) with a frequency of 0.75 Hz. Both stimuli are precisely aligned with the peak phase of the delta wave. In this embodiment, this phase-synchronized stimulation can effectively enhance the slow wave amplitude during deep sleep, thereby consolidating memory and promoting the clearance of metabolic waste from the brain. Furthermore, the output effect of this step S6 (such as the duration of N3 phase and changes in slow wave amplitude) is also recorded for the assessment of the "deep sleep deficiency path" in the weekly macro-loop (see description below).

[0050] S7: Implement a dual silencing protection strategy during rapid eye movement (REM) phase.

[0051] Rapid eye movement (REM) sleep is a crucial stage for dreaming and mood regulation during sleep. Any external stimulus can disrupt sleep continuity and even trigger nightmares. Therefore, when the real-time sleep staging engine 2 determines two consecutive staging windows (a total of 64 seconds) to be in the REM stage, the system immediately triggers a dual-silence protection mechanism. This dual-silence protection strategy, which only executes after two consecutive staging windows have been determined to be in the REM stage, avoids incorrect policy execution due to misjudgment of sleep staging.

[0052] Specifically, the physiological-psychological temporal coupling intervention module 6 outputs a forced zeroing and locking command, requiring the multimodal execution module 7 to set all physical stimulus output channels to zero and set a software lock flag. Thereafter, even if a single window is mistakenly identified as a non-REM period, the lock flag remains in place. Only after three consecutive phase windows (96 seconds) are identified as non-REM periods will the lock be released, restoring normal stimulus output. In this embodiment, this mechanism ensures, from both hardware and software perspectives, that the REM period is absolutely unaffected by any physical stimuli, thereby protecting the continuity of sleep and the physiological function of the REM period. Furthermore, if user awakening is detected during the lock period, the system will record the event as "REM period interruption" for subsequent analysis.

[0053] Since cognitive guidance is not initiated in the N2, N3, and REM stages, Table 2 only summarizes the characteristics of physiological stimulation in these three stages. Although cognitive guidance is not initiated in these three stages, slow-wave phase synchronization enhancement is performed during deep sleep to strengthen slow-wave activity, thereby promoting explicit memory consolidation. Double silencing protection is used during the REM stage to ensure the integrity of the REM stage, thereby promoting procedural memory and emotion regulation. This helps to automatically recall skills learned by the user during the day (such as diaphragmatic breathing, attention shifting, and cognitive reappraisal) at night (which is achieved through sleep-dependent memory consolidation).

[0054] Table 2 Summary of physiological stimuli during N2, N3 and REM phases

[0055] S8: Implement a smooth awakening strategy during the awakening period.

[0056] When the real-time sleep staging engine 2 determines that the user is in the wake phase, the system executes a smooth wake-up strategy. Unlike the physical de-stimulation during the sleep phase, waking up during the wake phase requires simulating the natural sunrise process to avoid sudden high-intensity stimulation that could startle or disturb the user.

[0057] Specifically, around the user's set wake-up time (usually within 30 minutes before the alarm), the multimodal execution module 7 controls the white LED array, gradually changing its color temperature from 2700K to 5000K, and linearly increasing its brightness to 200 lux within 10 minutes. Simultaneously, bone conduction speakers output natural ambient sounds (such as birdsong or a babbling brook), with the volume slowly increasing from 20dB to 45dB within 30 seconds. This gradual audio-visual stimulation smoothly transitions the user from sleep to wakefulness, enhancing the morning wake-up experience. During the awakening phase, cognitive guidance establishes pathways through the prefrontal cortex, achieving positive reinforcement and behavioral feedback; physiological stimulation establishes pathways through the suprachiasmatic nucleus (based on circadian rhythms, unlike the brainstem pathway established in the N1 stage) to the autonomic nervous system.

[0058] S9: Executes the gradual protocol during sleep stage transitions.

[0059] During a night's sleep, a user's sleep stages will switch multiple times. To avoid micro-awakening or sleep interruption caused by sudden changes in stimulation parameters, the system has a preset gradual transition protocol for stage switching.

[0060] Specifically, when the system detects that the current sleep stage is about to end, it linearly reduces all physical stimulus parameters output in the previous stage to zero within 10 seconds, prohibiting direct cut-off of output. When entering a new sleep stage, the system linearly increases the physical stimulus parameters required for the new stage from zero to the target baseline value within 5 seconds, prohibiting direct full-amplitude output. Thus, this stage-switching gradual protocol ensures the smoothness of intervention throughout the sleep cycle. This step S9 is typically applied between several sleep stages outside of REM sleep.

[0061] In addition, if a switch between REM and non-REM phases is detected, the system will prioritize the dual silencing protection strategy for REM phases (i.e., forced zeroing and locking), during which the phase switching gradient protocol will be suspended until the REM phase lock is released before resuming the gradient protocol.

[0062] In the above embodiments, preferably, the adaptive sleep regulation method further includes: S10: Day and night collaborative dual closed-loop adaptive iteration.

[0063] Understandably, the steps S1 to S9 described above constitute a "micro-loop" for real-time nighttime control, executed once every 32 seconds. However, users may develop stimulation tolerance or physiological changes during long-term use. Therefore, in the first embodiment of this invention, a morning macro-loop and a weekly macro-loop are also established to achieve daily and weekly parameter iteration.

[0064] Specifically, the execution process of the morning macro loop and the weekly macro loop is as follows: S101: Morning Macro Closed Loop.

[0065] After a user wakes up each day, the system calculates a personalized sleep quality index (SQI) based on the user's sleep data throughout the night (total sleep time, sleep latency, deep sleep percentage, REM sleep percentage, number of awakenings including micro-awakenings, etc.).

[0066] The SQI calculation formula is: SQI = w1 × sleep efficiency + w2 × (1 - sleep latency / personal baseline) + w3 × N3 phase percentage + w4 × (1 - number of nighttime awakenings / personal baseline) + w5 × subjective sleep score; where the weights w1 to w5 are personalized based on the user's historical data using the ridge regression algorithm, and the initial weights are the industry-standard benchmark values.

[0067] The system displays sleep reports to users via a mobile app, while dynamically adjusting the intervention baseline parameters for the night's micro-loop, such as the sleep onset acoustic-optical sweep rate, CES stimulus intensity baseline, and micro-arousal trigger threshold. These adjusted parameters will be used in steps S5 through S9 that night.

[0068] S102: Weekly macro closed loop.

[0069] Every week, the system aggregates all sleep data for the week, assesses the intervention's effectiveness, and resets the treatment plan based on five standardized pathways. Specifically, these five pathways include: ① Residual sleep onset difficulties (sleep latency still >30 minutes), corresponding to adjustments to sleep onset stimulation parameters next week (e.g., increase CES intensity by 10%, decrease the onset frequency of sound and light-induced stimulation by 2Hz); ② New early awakening pathway (morning awakening time >30 minutes earlier than the target), corresponding to starting morning light stimulation 15 minutes earlier next week; ③ The path of insufficient deep sleep (the proportion of N3 stage is lower than the age-matched baseline value) corresponds to a 10% increase in the intensity of slow-wave enhancement stimulation during deep sleep next week; ④ Insufficient compliance path (weekly average wearing rate <80%), corresponding to reducing the difficulty of training tasks and lowering the positive reinforcement trigger threshold by 5% next week; ⑤ Level transition path (if the insomnia severity score is lower than the current level threshold for 4 consecutive weeks and the sleep index returns to normal), the corresponding level will be automatically downgraded to a lower intervention level, and the stimulation intensity and training difficulty will be gradually reduced.

[0070] It is understood that the first embodiment of the present invention uses a three-layer nested adaptive architecture of "nighttime micro-loop - morning macro-loop - weekly macro-loop" to enable the system to maintain the intervention effect over a long period of time and avoid tolerance issues.

[0071] In the above embodiments, preferably, the adaptive sleep regulation method further includes: S20: Positive reinforcement and decoupling.

[0072] like Figure 3 As shown, in order to improve user compliance and gradually reduce device dependence, the first embodiment of this invention also introduces a positive reinforcement and decoupling mechanism. This mechanism is triggered after the morning macro loop execution and after daytime training is completed.

[0073] First, the system identifies success events. Success events are divided into two categories: (1) Physiological level, namely, the sleep quality index (SQI) of the night is not lower than the 80th percentile of the user's historical average, or is ≥15% higher than the same period last week; (2) Cognitive level, that is, at least one relaxation training session was completed the day before, and the heart rate variability (HRV) increased by ≥10% from the baseline during the training. As long as any one of these conditions is met, it is considered a successful event.

[0074] Upon recognizing a successful event, the system triggers dual-channel reinforcement feedback (i.e., physiological reinforcement and cognitive reinforcement). Specifically, the physiological reinforcement channel outputs gentle vibration pulses (lasting 1 second, frequency 80Hz, intensity 50% of the arousal stimulus) via the linear vibration motor in the multimodal execution module 7. The cognitive reinforcement channel displays positive verbal and visual feedback through a mobile app and unlocks corresponding sleep achievement badges. If successful events occur simultaneously at both levels, enhanced feedback is triggered, unlocking additional advanced cognitive behavioral therapy content.

[0075] Furthermore, the first embodiment of this invention also incorporates a decoupling design. Specifically, when the system triggers successful physiological events for three consecutive days, it determines that the user has developed a certain level of autonomous sleep regulation ability. At this point, a decoupling operation is automatically executed: the baseline intensity of physical stimulation in the micro-loop for that night is lowered by 5% (relative to the current baseline), with the cumulative reduction not exceeding 50% of the initial baseline. Simultaneously, the trigger threshold for positive reinforcement is also fine-tuned, and autonomous regulation training content in device-free scenarios is added. This design strengthens the user's positive behavior while gradually reducing external intervention, guiding the user to gradually break free from dependence on devices.

[0076] Second Embodiment like Figure 4 As shown, the second embodiment of the present invention provides a physiological-psychological integrated adaptive sleep regulation system for executing the adaptive sleep regulation method in the first embodiment. Specifically, the adaptive sleep regulation system includes a multimodal perception module 1, a real-time sleep staging engine 2, a cortical arousal level detection unit 3, a micro-arousal detection unit 4, a specific intervention rule base 5, a temporal coupling intervention module 6, a multimodal execution module 7, an adaptive decision engine 8, and a positive reinforcement and dedependency coupling module 9. The multimodal sensing module 1 is used to acquire the user's electroencephalogram (EEG), cardiovascular, and body movement signals in real time. Specifically, it includes an EEG acquisition unit 11 (single-channel frontal dry electrode), a cardiovascular signal acquisition unit 12 (ear clip-on PPG sensor), and a behavioral signal acquisition unit 13 (triaxial accelerometer). All signal acquisition is completed at the device edge, and the raw data is not transmitted back to the cloud.

[0077] The real-time sleep staging engine 2, connected to the multimodal perception module 1, is used to determine the user's current sleep stage based on EEG signals. This engine employs a CNN-BiLSTM hybrid deep learning model, outputting five-category sleep staging results (wakefulness, sleep onset, light sleep, deep sleep, and REM sleep) within a 32-second time window. The model is compressed to ≤5MB through knowledge distillation, and the edge inference latency is ≤100ms (see step S2 of the first embodiment for specific implementation details).

[0078] The cortical arousal level detection unit 3 is connected to the multimodal perception module 1 and is used to assess the user's cortical arousal level in real time based on the Beta wave power in the electroencephalogram (EEG) signal, with 120% of the resting baseline as the high arousal threshold (see step S3 of the first embodiment for specific determination rules).

[0079] The micro-awakening detection unit 4 is connected to the multimodal sensing module 1 and is used to detect micro-awakening events based on cardiovascular signals and / or body movement signals. The determination rules are as described in step S5 of the first embodiment.

[0080] The sleep stage-specific intervention rule base 5 stores the range of physical stimulation parameters, contraindications, and timing pairing strategies corresponding to each sleep stage (physiological downgrading during sleep onset, subthreshold maintenance during light sleep, slow wave phase synchronization during deep sleep, double silence protection during REM sleep, and smooth awakening during wakefulness).

[0081] The sleep-stage-specific intervention rule base 5 is one of the core modules of the system. This rule base stores the range of physical stimulus parameters corresponding to each sleep stage, the contraindications for each stage, and the corresponding temporal pairing strategies (such as...). Figure 5 (As shown). Specific details are shown in Table 3 below (corresponding to steps S4-S8 of the first embodiment above): Table 3 Taboo rules for each stage and corresponding time-series pairing strategies for each stage

[0082] In addition, the specific intervention rule base 5 also has a built-in phase switching gradual protocol, the details of which can be found in step S9 of the first embodiment.

[0083] The physiological-psychological temporal coupling intervention module 6, connected to the real-time sleep staging engine 2 and the cortical arousal level detection unit 3, is used to execute intervention decisions based on the user's sleep stage and cortical arousal level, according to the temporal pairing strategies corresponding to each stage in the sleep stage-specific intervention rule base 5. Specifically, during the sleep onset period, temporal coupling decisions are executed (including full physical stimulation output in a high arousal state, cognitive guidance initiation and coordination maintenance mode switching after low arousal stabilization, and rollback protection mechanism). During the light sleep, deep sleep, REM sleep, and wakefulness periods, corresponding control instructions are output according to the corresponding strategies in the rule base.

[0084] The multimodal execution module 7 is used to output physical stimuli and / or cognitive guidance content according to the instructions of the physiological-psychological temporal coupling intervention module 6. Its hardware components include a bone conduction speaker, a red / white LED array, a CES microcurrent output module, a graphene temperature-controlled heating element, and a linear vibration motor, which can be integrated into a head-mounted eye mask.

[0085] The day-night collaborative dual-closed-loop adaptive decision engine 8 is used to execute the morning macro-closed loop and weekly macro-closed loop described in step S10 of the first embodiment. It calculates the personalized sleep quality index (SQI) based on the whole night's sleep data and dynamically adjusts the intervention baseline parameters. It also resets the overall intervention plan for the next week through five standardized paths (residual difficulty falling asleep, new early awakening, insufficient deep sleep, insufficient compliance, and hierarchical transition) (see the first embodiment for details).

[0086] The positive reinforcement and decoupling module 9 is used to perform the positive reinforcement and decoupling described in step S20 of the first embodiment, including dual-pathway successful event recognition, dual-channel reinforcement feedback, and baseline downregulation of stimulus intensity after continuous achievement of targets (see the first embodiment for details).

[0087] It is understood that the functions and collaborative workflows of the above modules are only one specific implementation of the method of the first embodiment. In other embodiments, the functions and collaborative workflows of each module can be adaptively adjusted according to actual needs to achieve the adaptive sleep regulation method described in the first embodiment, which is not specifically limited here.

[0088] Third Embodiment like Figure 6 As shown, based on the first embodiment described above, the third embodiment of the present invention also provides a skill generalization training method for the aforementioned adaptive sleep regulation system. This method helps users establish autonomous relaxation ability in device-free scenarios through breathing-stimulus coupling anchoring, and specifically includes the following stages: I. Daytime Training Phase Users perform daytime relaxation exercises via a mobile app. The system extracts the user's breathing rhythm in real time using PPG signals and calculates the real-time respiratory rate f. The following respiratory coupling algorithm is executed: ① The stimulation carrier frequency is set to 3×f, which is completely synchronized with the respiratory rhythm; ②At the end of the inspiratory phase (200ms before the peak of the respiratory wave), the stimulus intensity increases linearly by 8%; ③ At the end of the expiratory phase (200ms before the respiratory trough), the stimulation intensity is reset to the baseline value; ④ Synchronously output diaphragmatic breathing guidance voice that matches the breathing rhythm, guiding the user to stabilize the breathing rate at a relaxation frequency of 6-8 breaths / minute.

[0089] II. Anchoring Phase The coupling parameters (stimulation carrier frequency = 3 × respiratory rate, 8% increase at the end of inspiration, and reset at the end of expiration), guiding voice content, and stimulation mode used during the daytime training phase were set to be completely consistent with the intervention plan executed by the system during the nighttime sleep period.

[0090] In this embodiment, this anchoring strengthens the conditioned reflex of "breathing rhythm adjustment → stimulus change → relaxation experience".

[0091] III. Migration and Application Phase After 7 to 14 days of continuous training, users can achieve autonomous sleep regulation by adjusting their breathing to the frequency learned during training (6 to 8 breaths per minute) in a device-free environment (such as without using an eye mask or the app) and evoking the same relaxation experience as during the training period.

[0092] Furthermore, this method supports stratified adaptation based on the Insomnia Severity Index (ISI) and the Pittsburgh Sleep Quality Index (PSQI): users are categorized into healthy individuals (ISI<7), subclinical individuals (7≤ISI<14), individuals with mild to moderate insomnia (14≤ISI<21), and individuals with severe insomnia (ISI≥21). Differentiated training difficulty, duration, and frequency are then configured for each level. For example, individuals with mild to moderate insomnia receive 20 minutes of training three times daily to promote symptom improvement and functional recovery; while individuals with severe insomnia receive training with a reduced difficulty level, focusing on symptom control and better access to medical care.

[0093] In summary, the physiological-psychological integrated adaptive sleep regulation system and method provided in this invention can be integrated into various forms such as head-mounted eye masks, in-ear headphones, smart bracelets, or smart mattresses. It is suitable for sleep optimization in healthy individuals, early intervention in subclinical populations, adjunctive treatment for insomnia patients, and various application scenarios including shift workers, adolescents, perioperative patients, Alzheimer's disease prevention, and post-traumatic stress disorder. This invention can significantly improve the accuracy and effectiveness of sleep regulation, reduce user dependence on devices, and has good industrial practical value and commercial prospects.

[0094] It should be noted that the above embodiments are merely illustrative examples. The technical solutions of each embodiment can be combined, and all are within the protection scope of this invention.

[0095] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0096] The above provides a detailed description of the physiological-psychological integrated adaptive sleep regulation method and system provided by this invention. Any obvious modifications made by those skilled in the art without departing from the essence of this invention will constitute an infringement of the patent rights of this invention and will incur corresponding legal liability.

Claims

1. A physiological-psychological integrated adaptive sleep regulation method, characterized in that... Includes the following steps: Real-time acquisition of users' brainwave signals; The user's current sleep stage is determined based on the EEG signals, which include the wakefulness stage, the sleep onset stage, the light sleep stage, the deep sleep stage, and the REM sleep stage; and the user's cortical arousal level is assessed in real time based on the Beta wave power in the EEG signals. When the user is in the sleep stage, a physiological downgrade-first strategy is executed based on the cortical arousal level: when the cortical arousal level is higher than a preset threshold, a full physical stimulation command is output; when the cortical arousal level drops below the preset threshold and remains below it for a preset time, a cognitive guidance initiation command is output, and the physical stimulation is switched to a coordinated maintenance mode; the coordinated maintenance mode means that the output intensity of acoustic and light stimulation is reduced to below the user's perception threshold, and electrical stimulation is maintained at a low intensity level below the full output during the sleep stage; furthermore, if the cortical arousal level rises above the preset threshold again during the cognitive guidance output process, the cognitive guidance output is immediately paused and the full physical stimulation output is resumed. When the user is in a light sleep period, a preset subthreshold maintenance strategy is executed: cognitive guidance is stopped, and short-term enhanced stimulation is triggered only when micro-awakening is detected; When the user is in deep sleep, a preset slow-wave phase synchronization strategy is executed: pulse stimulation synchronized with the slow oscillation phase of the brain is output, thereby enhancing slow-wave activity; When the user is in the REM (Rapid Eye Movement) phase, a preset dual-silence protection strategy is executed: after two consecutive phase windows confirm the entry into the REM phase, all physical stimulus outputs are forced to zero and locked; after three consecutive phase windows confirm the exit from the REM phase, the lock is released. When the user is in the waking period, a preset smooth wake-up strategy is executed: simulating the gradual change of sunrise light and the gradual output of natural ambient sound.

2. The adaptive sleep regulation method as described in claim 1, characterized in that... Also includes: When a sleep phase transition is detected, the phase transition gradient protocol is executed; The phase switching gradual protocol includes: the stimulation parameters exiting the previous phase linearly return to zero within 10 seconds, and the stimulation parameters entering the new phase linearly increase to the target baseline value within 5 seconds from zero; and when a switch between the rapid eye movement (REM) phase and the non-REM phase is detected, the dual silencing protection strategy of the REM phase is executed first, and the phase switching gradual protocol is paused.

3. The adaptive sleep regulation method as described in claim 1, characterized in that... Also includes: Collect the user's cardiovascular signals and / or body movement signals, and detect micro-arousal events based on the cardiovascular signals and / or body movement signals; When a micro-awakening event is detected during light sleep, a short-term enhanced physical stimulus output is triggered, which automatically returns to the subthreshold level after a preset time.

4. The adaptive sleep regulation method as described in claim 1, characterized in that... It also includes a day-night collaborative dual-closed-loop adaptive iterative step: A nighttime micro-loop is implemented to drive the adjustment of stimulation parameters based on real-time sleep stage results; The morning macro loop is executed, the sleep quality index is calculated based on the overnight sleep data, and the intervention baseline parameters of the nighttime micro loop are dynamically adjusted for that night. A weekly macro-loop is implemented, and based on the sleep data of the whole week, the overall intervention plan for the following week is reset through multiple standardized paths.

5. The adaptive sleep regulation method as described in claim 4, characterized in that... The standardized pathways include at least: residual sleep onset difficulty pathways, new early awakening pathways, insufficient deep sleep pathways, insufficient compliance pathways, and hierarchical transition pathways; Among them, the residual difficulty falling asleep path corresponds to the scheme reset of adjusting the stimulation parameters during the sleep onset period, the newly emerging early awakening path corresponds to the scheme reset of adjusting the morning light stimulation time, the insufficient deep sleep path corresponds to the scheme reset of adjusting the intensity of slow wave enhancement stimulation during deep sleep, the insufficient compliance path corresponds to the scheme reset of adjusting the training difficulty and trigger threshold, and the hierarchical transition path corresponds to the scheme reset of reducing the overall stimulation intensity.

6. The adaptive sleep regulation method as described in claim 1, characterized in that... It also includes positive reinforcement and decoupling steps: When any successful event is detected, dual-channel reinforcement feedback is triggered, including: the physiological reinforcement channel outputs a gentle vibration pulse, and the cognitive reinforcement channel displays positive voice and visual feedback; the successful events include: the nighttime sleep quality index reaching or improving at the physiological level, and the completion of daytime relaxation training and improvement of heart rate variability at the cognitive level. When a successful event is recognized simultaneously at both the physiological and cognitive levels, an enhanced feedback is triggered. When a physiological success event is triggered for three consecutive days, the intensity baseline of the physical stimulus is automatically reduced by 5%, with the cumulative reduction not exceeding 50% of the initial baseline.

7. The adaptive sleep regulation method as described in claim 1, characterized in that: The preset threshold is 120% of the user's resting-state EEG Beta wave baseline; when the user's EEG Beta wave relative power is higher than 120% of the baseline value, it is determined to be a high arousal state; when the user's EEG Beta wave relative power is lower than 120% of the baseline value and lasts for 10 seconds, it is determined to be a low arousal stable state.

8. The adaptive sleep regulation method as described in claim 1, characterized in that... Determining the user's current sleep stage based on the electroencephalogram (EEG) signals specifically includes: The EEG signals are input into a pre-trained CNN-BiLSTM hybrid deep learning model in a 32-second time window; Time-domain and frequency-domain features are extracted using a three-layer convolutional neural network; The time-domain and frequency-domain features are input into a two-layer bidirectional long short-term memory network to extract the temporal context features of sleep stage changes; Based on the temporal context features, a five-category sleep stage result is output, which is one of the following: wakefulness, sleep onset, light sleep, deep sleep, and REM sleep.

9. A physiological-psychological integrated adaptive sleep regulation system, used to implement the adaptive sleep regulation method according to any one of claims 1 to 8, characterized in that... include: A multimodal sensing module is used to collect the user's electroencephalogram (EEG) signals in real time; The real-time sleep staging engine, connected to the multimodal perception module, is used to determine the user's current sleep stage based on the EEG signals. The sleep stages include wakefulness, sleep onset, light sleep, deep sleep, and REM sleep. A cortical arousal level detection unit, connected to the multimodal perception module, is used to assess the user's cortical arousal level in real time based on the Beta wave power in the electroencephalogram (EEG) signal. A sleep stage-specific intervention rule base stores the range of physical stimulation parameters corresponding to each sleep stage, the contraindication rules for each stage, and the time-series pairing strategies for each stage. The time-series pairing strategies include: a physiological downgrade strategy for the sleep onset period, a subthreshold maintenance strategy for the light sleep period, a slow wave enhancement strategy for the deep sleep period, a dual silence protection strategy for the REM sleep period, and a smooth awakening strategy for the wakefulness period. The physiological-psychological temporal coupling intervention module connects the real-time sleep staging engine and the cortical arousal level detection unit. It is used to execute intervention decisions based on the user's sleep stage and cortical arousal level, according to the temporal pairing strategy corresponding to each stage in the sleep stage-specific intervention rule base. The multimodal execution module is used to output physical stimuli and / or cognitive guidance content according to instructions.

Images