A method of sleep state monitoring and electronic device

Abstract

The present application relates to the technical fields of physiological signal monitoring and wearable electronic devices. The present application discloses a sleep state monitoring method and an electronic device. Local PPG signals and contact loads at two detection positions are collected to obtain a candidate sleep state conversion. In response to the candidate conversion or a conflict between double-zone evidence, a differential micro-load component is controlled to increase and decrease the loads at the two positions and to perform reverse adjustment. Verification data is obtained within the local and total load safety intervals, a double-zone local load response relationship is established, and the PPG features of the candidate conversion window are corrected. Common physiological components and contact disturbance components are extracted through pulse pairing and direction classification, the candidate conversion is arbitrated, and a sleep state monitoring result is obtained. This scheme can reduce the influence of PPG artifacts caused by factors such as changes in contact load, wearing and repositioning on sleep state judgment, and also takes into account the wearing safety during the verification process.

Description

Technical Field

[0001] This invention relates to the field of sleep monitoring technology, specifically to a method and electronic device for monitoring sleep states. Background Technology

[0002] Wearable sleep monitoring devices typically collect signals such as photoplethysmography (PPG), heart rate, heart rate variability, body movement, and temperature in the nighttime wearing environment. They extract features such as pulse interval, waveform amplitude, respiratory modulation, and activity intensity from a continuous sampling window, and then use threshold judgment, finite state machine, or other signal processing methods to form state outputs such as wakefulness, sleep transition, stable sleep, or short-term awakening, in order to achieve continuous monitoring of the wearer's sleep process.

[0003] Swelling of the fingers, clenching of the fist, pressing of the hand, turning over, or device repositioning during sleep can alter the contact load between the optical sensor and the skin, causing simultaneous changes in the optical path, signal-to-noise ratio, and local microvascular pressure. Existing processing typically relies on the relationship that optical changes in the sampling window primarily reflect physiological changes, or deletes the window and selects channels with higher signal quality when pressure is abnormal. However, when changes in contact load are similar to actual cardiovascular or autonomic nervous system changes in time and waveform, the above methods struggle to determine their origins, leading to data gaps near candidate transitions or interpreting contact perturbations as changes in sleep state, ultimately resulting in erroneous state transition results. Summary of the Invention

[0004] In view of the above-mentioned problems in the prior art, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for monitoring sleep states, wherein the local PPG signal includes a first PPG signal and a second PPG signal, the local contact load includes a first actual contact load and a second actual contact load, the detection position includes a first detection position and a second detection position, and the dual-zone local load response relationship includes a first local load response relationship and a second local load response relationship; the first PPG signal and the first actual contact load originate from the first detection position, and the second PPG signal and the second actual contact load originate from the second detection position; the first local load response relationship belongs to the first detection position, and the second local load response relationship belongs to the second detection position; the sum of the first actual contact load and the second actual contact load is the total contact load; the method includes: S1. Acquire the local PPG signal and the local contact load, and obtain candidate sleep state transitions and candidate transition windows based on the local PPG signal; S2. In response to the candidate sleep state transition or the conflict of transition evidence between the first PPG signal and the second PPG signal, control the differential microload component to perform differential microload verification. The differential microload verification adjusts the local contact load by increasing and decreasing it in reverse according to the feedback of the local contact load and the total contact load, limiting the local contact load to a preset local safety range and the total contact load to a preset total safety range. The local PPG signal and the local contact load during the adjustment period are collected to obtain microload verification data. S3. Establish the dual-zone local load response relationship based on the microload verification data, and use the dual-zone local load response relationship and the local contact load in the candidate conversion window to correct the PPG features in the candidate conversion window to obtain the dual-zone load normalized PPG feature sequence. S4. Perform pulse pairing and direction classification on the dual-zone load normalized PPG feature sequence, extract common physiological components and contact perturbation components, arbitrate the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtain the arbitration result, and obtain the sleep state monitoring result from the arbitration result.

[0005] Optionally, obtaining the candidate sleep state transition and candidate transition window based on the local PPG signal includes: Pulse interval trends, relative changes in peripheral vascular tension, and PPG respiratory modulation were extracted from the first PPG signal and the second PPG signal. Using the pulse interval trend, the relative change in peripheral vascular tension, and the PPG respiratory modulation as inputs, the candidate sleep state transitions are obtained using a finite state machine with an entry threshold, an exit threshold, and a minimum duration, and the continuous sampling intervals that meet the transition conditions are determined as the candidate transition windows.

[0006] Optionally, the differential microload verification includes: The neutral phase, the first differential phase, the second differential phase, the second differential phase, the first differential phase, and the neutral phase are executed sequentially. In the first differential phase, the first actual contact load is increased and the second actual contact load is decreased; in the second differential phase, the first actual contact load is decreased and the second actual contact load is increased; each phase continues until a complete pulse is detected. PPG sampling values ​​and actual contact load sampling values ​​are collected in each phase. The PPG sampling values ​​and actual contact load sampling values ​​of each phase are arranged in time to obtain the microload verification data.

[0007] Optionally, the differential microload verification specifically includes: Read the local load allowable range and the total load allowable range, and adjust the driving amount of the differential micro-load component according to the feedback of the first actual contact load and the second actual contact load, so that the first actual contact load and the second actual contact load are respectively within the local load allowable range, and the sum of the first actual contact load and the second actual contact load is within the total load allowable range; Inertial signals are collected using inertial sensors. When any of the local contact loads exceeds the local load allowable range, or the sum of the first actual contact load and the second actual contact load exceeds the total load allowable range, or the inertial signal indicates overall slippage, or no complete pulse is detected, the differential micro-load component is controlled to return to a neutral state and stop using the micro-load verification data.

[0008] Optionally, the differential micro-load assembly includes a piezoelectric bicrystalline beam, a first skin-contact pad, and a second skin-contact pad. The two ends of the piezoelectric bicrystalline beam are mechanically connected to the first skin-contact pad and the second skin-contact pad, respectively. The driving polarity switching of the piezoelectric bicrystalline beam causes the normal displacement directions of the first skin-contact pad and the second skin-contact pad to be interchanged, so as to adjust the first actual contact load and the second actual contact load, respectively.

[0009] Optionally, the differential micro-load assembly includes an eccentric cam, a rocker arm, a neutral return elastic element, a first skin pad, and a second skin pad. The rocker arm uses the housing of the electronic device as a fulcrum. The first skin pad is mechanically connected to the first output end of the rocker arm, and the second skin pad is mechanically connected to the second output end of the rocker arm. The eccentric cam drives the rocker arm through mechanical contact, causing the first output end and the second output end to move along opposite skin normals.

[0010] Optionally, the step of establishing the dual-zone local load response relationship based on the micro-load verification data, and using the dual-zone local load response relationship and the local contact load within the candidate conversion window to correct the PPG features within the candidate conversion window, to obtain the dual-zone load normalized PPG feature sequence, includes: Each micro-load phase is composed of data from the micro-load verification data that have the same direction of increase or decrease in local contact load; Normalized PPG features and actual contact load changes are extracted from each micro-load phase, and the local load response relationship of the dual-zone is established based on the normalized PPG features and the actual contact load changes. The static load ratio, the optical DC component, and the electronic device orientation acquired by the inertial sensor are compared between the end of the candidate conversion window and the start time of the differential microload verification. When all differences are within the continuity threshold, a contact geometric continuity determination is obtained. The difference between the first actual contact load within the candidate conversion window and the first actual contact load before the start time of the differential micro-load verification is calculated to obtain the first compensation load. The difference between the second actual contact load within the candidate conversion window and the second actual contact load before the start time of the differential micro-load verification is calculated to obtain the second compensation load. When the result of the contact geometry continuity determination is continuous, and both the first compensation load and the second compensation load are within the load change range of the differential micro-load verification, the PPG features in the candidate conversion window are corrected using the dual-zone local load response relationship, the first compensation load, and the second compensation load to obtain the dual-zone load normalized PPG feature sequence.

[0011] Optionally, the first local load response relationship is the median of the ratio of the normalized PPG characteristic change of the first PPG signal to the change of the first actual contact load. The second local load response relationship is the median of the ratio of the normalized PPG characteristic change of the second PPG signal to the change of the second actual contact load; Actual contact load changes whose absolute value is less than the preset load resolution threshold are not included in the median calculation.

[0012] Optionally, the step of performing pulse pairing and direction classification on the dual-zone load-normalized PPG feature sequence, extracting common physiological components and contact perturbation components, arbitrating the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtaining an arbitration result, and obtaining sleep state monitoring results from the arbitration result includes: The common physiological component is a feature change that is paired with the pulse time, has the same direction, and exceeds the measurement uncertainty. The contact disturbance component is a feature change with opposite direction, a feature change that exists only at the first detection position, or a feature change that exists only at the second detection position. Based on the contact disturbance component, the presence status of the transition evidence for the candidate sleep state transition in the local PPG signal and the dual-zone load normalized PPG feature sequence is compared to obtain the determination of the disappearance of transition evidence. When the common physiological components satisfy the migration conditions for the candidate sleep state transition, the arbitration result is acceptance; otherwise, when the result of the determination of the disappearance of the transition evidence is disappearance, the arbitration result is rejection; otherwise, the arbitration result is indeterminate. The sleep state monitoring results are obtained based on the arbitration result.

[0013] Secondly, the present invention provides an electronic device for monitoring sleep states, comprising a wearable housing; a first measuring component including a first skin-contact pad, a first PPG sensor, and a first load sensor, disposed at a first detection position; a second measuring component including a second skin-contact pad, a second PPG sensor, and a second load sensor, disposed at a second detection position; the first measuring component and the second measuring component are disposed at intervals on the same skin-contact side of the wearable housing; a passive support pad disposed on opposite sides of the first measuring component and the second measuring component; a differential micro-load component mechanically connected to the first skin-contact pad and the second skin-contact pad respectively, for moving the first skin-contact pad and the second skin-contact pad along opposite skin normals; an inertial sensor; a synchronous sampling front end electrically connected to the first PPG sensor, the second PPG sensor, the first load sensor, the second load sensor, and the inertial sensor respectively; and a controller electrically connected to the synchronous sampling front end and the differential micro-load component respectively, for executing the aforementioned method.

[0014] Beneficial effects The technical solution provided by this invention has the following beneficial effects: 1. Local PPG signals and actual contact loads are simultaneously acquired at two detection locations on the wearable device. Differential microload verification is performed only when there is conflicting evidence of candidate sleep state transitions or dual-zone transitions. The local load response relationship of the dual zones under the current contact geometry is established based on the measured load. After correcting the candidate transition window using this relationship, the common physiological component and contact perturbation component are extracted based on the change direction of the paired pulse. This can separate the optical and microvascular perturbations that change with local load from the physiological changes that occur in both zones. This allows candidate transitions dominated by contact perturbations to be rejected, while candidate transitions supported by common physiological changes to be accepted. In cases of insufficient evidence, uncertainty is output, thereby reducing the interpretation of PPG changes caused by hand pressure, repositioning, or nocturnal edema as sleep state transitions. At the same time, it avoids data gaps caused by directly deleting abnormal windows.

[0015] 2. By adjusting the two skin pads on the same skin-contacting side by increasing and decreasing their pressure and in the opposite direction, the local contact load is transferred under the condition that the passive support pad stabilizes the device position. Based on the feedback from the two load sensors, the local contact load and the total contact load are limited simultaneously. Combined with event-triggered drive, inertial signal gating, complete pulse detection, abnormal return to center, and stopping the use of invalid data, short-term verification can be completed without continuously tightening the device, limiting the changes in local pressure and overall restraint during the verification process, and reducing the drive duty cycle of the differential micro-load component. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the wearing of the smart ring and the contact scenario of the dual detection areas provided by the present invention; Figure 2 This invention provides a schematic diagram of the steps involved in a method for monitoring sleep states. Figure 3 This invention provides an overall flowchart of a method for monitoring sleep states. Figure 4 A schematic diagram of the structure of an electronic device for monitoring sleep states provided by the present invention; Figure 5 This is a schematic diagram of the piezoelectric bicrystalline beam differential micro-load assembly structure provided by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This embodiment uses a smart ring as the primary application carrier. The smart ring is worn on the wearer's finger during sleep. A first and second measuring component are spaced apart along the finger's axis on the same skin-contact side of the wearable shell, contacting the skin at the first and second detection positions, respectively. A passive support pad is positioned on the opposite side, with a contact area larger than any single skin-contact pad, serving to form a stable fulcrum. The differential micro-load component only transfers local loads between two adjacent detection positions, without verification through continuous tightening of the entire ring. The smart ring can upload sleep state monitoring results to a paired terminal, but synchronous acquisition, drive control, safety abort, and return-to-center control are all completed locally on the smart ring; communication interruptions do not affect these local functions.

[0019] The operating environment of this embodiment includes a battery-powered controller, memory, a synchronous sampling front-end, two PPG sensors, two load sensors, an inertial sensor, a differential microload driver, and an optional wireless communication unit. The memory stores program instructions, individual resting baselines, allowable local load ranges, allowable total load ranges, measurement uncertainty, a finite state machine parameter table, candidate transition evidence records, and valid local load response relationships. Each synchronous sampling frame contains at least a monotonically increasing timestamp, a detection location identifier, a PPG sample value, the actual contact load in Newtons, and the corresponding inertial data. As a non-limiting configuration, the two PPG channels and two load channels can be synchronously sampled at 200 Hz, and the three-axis inertial data can be sampled at 50 Hz. The controller provides a unified time base for optical emission, analog-to-digital conversion, and drive phase via a common timer; the sampling rate can also be adjusted according to pulse bandwidth, power budget, and sensor range.

[0020] This embodiment first provides a method for monitoring sleep states.

[0021] In some embodiments, the local PPG signal includes a first PPG signal and a second PPG signal, the local contact load includes a first actual contact load and a second actual contact load, the detection position includes a first detection position and a second detection position, and the dual-zone local load response relationship includes a first local load response relationship and a second local load response relationship; the first PPG signal and the first actual contact load originate from the first detection position, and the second PPG signal and the second actual contact load originate from the second detection position; the first local load response relationship belongs to the first detection position, and the second local load response relationship belongs to the second detection position; the sum of the first actual contact load and the second actual contact load is the total contact load; the method includes: S1. Acquire the local PPG signal and the local contact load, and obtain candidate sleep state transitions and candidate transition windows based on the local PPG signal; S2. In response to the candidate sleep state transition or the conflict of transition evidence between the first PPG signal and the second PPG signal, control the differential microload component to perform differential microload verification. The differential microload verification adjusts the local contact load by increasing and decreasing it in reverse according to the feedback of the local contact load and the total contact load, limiting the local contact load to a preset local safety range and the total contact load to a preset total safety range. The local PPG signal and the local contact load during the adjustment period are collected to obtain microload verification data. S3. Establish the dual-zone local load response relationship based on the microload verification data, and use the dual-zone local load response relationship and the local contact load in the candidate conversion window to correct the PPG features in the candidate conversion window to obtain the dual-zone load normalized PPG feature sequence. S4. Perform pulse pairing and direction classification on the dual-zone load normalized PPG feature sequence, extract common physiological components and contact perturbation components, arbitrate the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtain the arbitration result, and obtain the sleep state monitoring result from the arbitration result.

[0022] The four steps of the above method are as follows: Figure 2 As shown, the complete normal path, delayed path, aborted path, and uncertain path are as follows: Figure 3 As shown. This method does not directly interpret contact load as a physiological indicator, but instead uses short-term and limited differential load as a local identification input; the relationship obtained from this input only serves the PPG feature correction and sleep state transition arbitration under the current contact geometry.

[0023] For S1, specifically, after the controller starts, it first reads the zero point and range calibration parameters of the two load channels, and then reads the DC component and pulsation component of PPG at two locations. Continuous monitoring only begins when both actual contact loads are within the measurable contact area, at least one can continuously detect a pulse, and the total contact load is lower than the non-blocking wearing limit obtained from assembly calibration. If either load channel is saturated, neither can detect a pulse, or the total contact load exceeds the wearing limit, an abnormal wearing condition is output, the differential micro-load assembly is not activated, and the system waits for the load and optical signal to recover.

[0024] The controller performs DC drift removal, bandpass filtering, pulse peak and pulse origin detection on the local PPG signal, and forms a stroke feature record according to a continuous time window. The stroke feature record includes at least one or more of the following: pulse time, pulse interval, AC / DC ratio, normalized pulse amplitude, normalized pulse area, and rise time. When establishing an individual resting baseline, the controller selects a common baseline feature for the relative change in peripheral vascular tension at both detection locations that passes quality judgment, based on normalized pulse amplitude and normalized pulse area; if no common baseline feature exists, the evidence is marked as invalid. PPG respiratory modulation uses an envelope modulation approach: the controller also selects a common baseline feature that passes quality judgment at both detection locations, forming an amplitude envelope or area envelope arranged according to pulse time, and then calculates the proportion of respiratory band power of this envelope to the total variation power of the envelope; the selected baseline feature is not changed during a single candidate processing period. The pulse interval is only used to calculate the pulse interval trend and is not used as the PPG respiratory modulation input in this embodiment. If green light and infrared light channels are used, the channel quality can be judged within each detection location first, and then the same type of channel that has passed the quality judgment at both locations can be used for the above feature extraction. Instead of selecting a single optimal channel across detection locations to replace dual-zone processing, this approach is not used.

[0025] In some implementations, obtaining candidate sleep state transitions and candidate transition windows based on the local PPG signal includes: Pulse interval trends, relative changes in peripheral vascular tension, and PPG respiratory modulation were extracted from the first PPG signal and the second PPG signal. Using the pulse interval trend, the relative change in peripheral vascular tension, and the PPG respiratory modulation as inputs, the candidate sleep state transitions are obtained using a finite state machine with an entry threshold, an exit threshold, and a minimum duration, and the continuous sampling intervals that meet the transition conditions are determined as the candidate transition windows.

[0026] Specifically, the finite state machine includes states of wakefulness, sleep transition, stable sleep, suspected sleep disturbance or short-term awakening, and indeterminate states. The controller generates a ternary evidence vector for the first and second detection locations, consisting of pulse interval trend, relative change in peripheral vascular tension, and PPG respiratory modulation. The pulse interval trend is the robust linear slope of the effective pulse interval within the current analysis window; the relative change in peripheral vascular tension is the inverse of the relative difference between the normalized pulse amplitude or area median of the current window and the individual's resting baseline median; a positive value indicates a relative increase in tension, and this item is marked as invalid when the resting baseline is below the optical reliability lower limit; PPG respiratory modulation is the proportion of the power of the selected pulse amplitude or area envelope within the respiratory band to the total variation power of the envelope, and this item is marked as invalid when the total variation power is below the resting noise upper limit. All three inputs carry feature identifiers, basic feature identifiers, detection locations, time ranges, and quality indicators.

[0027] The finite state machine parameter table stores the direction sign, entry threshold, exit threshold, and minimum duration window for each of the three inputs for each target migration and each detection position. In an executable configuration, when transitioning from wakefulness to sleep, prolonged pulse interval, decreased peripheral vascular tone, and increased respiratory modulation stability are set as supporting directions; the opposite supporting directions are used when transitioning from stable sleep to suspected sleep disturbances or short-term awakenings. The controller first converts each input into supporting evidence according to the direction sign of the target migration, and then compares the corresponding thresholds. The entry threshold is taken as the robust high quantile of the corresponding supporting evidence in the individual's resting baseline, and the exit threshold is taken as the quantile closer to the baseline median to form hysteresis; when the baseline dispersion is lower than the minimum resolvable variation calibrated by the product, this minimum resolvable variation is used to limit the threshold spacing.

[0028] As a non-limiting example, the controller employs a 30-second analysis window and a 5-second sliding step. When at least two of the three supporting pieces of evidence for any target migration reach the entry threshold for three consecutive windows, a corresponding local candidate migration is generated. During the sleep transition state, the controller enters stable sleep when the same-direction condition continues to satisfy a preset stable duration window and there is no evidence in the opposite direction. The current candidate state is exited when at least two supporting pieces of evidence fall below their respective exit thresholds for two consecutive windows. If only one piece of evidence is valid, the quality indicators of the three pieces of evidence do not meet the minimum quantity, or the forward and reverse direction conditions at the same detection position are simultaneously met, the original state is maintained and the evidence is marked as insufficient. The candidate transition window for this local candidate migration is formed from the first frame that satisfies the migration conditions to the currently satisfied frame, and the controller simultaneously saves its preceding and following buffer data.

[0029] Each local candidate migration generates a local candidate transition evidence record. This record includes at least the candidate identifier, current state, single target state, migration direction, source detection location, feature identifiers involved in the judgment, basic feature identifiers used to generate each piece of evidence and their supporting directions, entry and exit thresholds for the corresponding detection location, threshold version, minimum duration, start and end times of the candidate transition window, and quality flags. The controller groups local candidate records with overlapping timeframes into the same verification event and assigns a unique verification event identifier to this event; local candidate records without overlapping timeframes form separate verification events. The time range of a verification event is from the earliest start to the latest end of its contained local candidate windows.

[0030] For the same verification event, the controller generates one or more candidate transition evidence records according to the following rules: When both the first and second detection positions satisfy the same target migration, the two local candidate records are merged into one candidate transition evidence record, and the temporal intersection of the two local candidate windows is determined as the candidate transition window of that record; when only one detection position satisfies the target migration, a candidate transition evidence record with that target state as the sole objective is generated, the local candidate window of that position is used as the candidate transition window, and a single-area evidence conflict is marked; when two detection positions satisfy different target migrations, two candidate transition evidence records are generated respectively, each record only saves its corresponding target state and its source position's local candidate window, and a direction conflict is marked; when a candidate migration coincides with a load mutation, the target state and candidate transition window of the original candidate transition evidence record are retained and a load mutation conflict flag is added, and no new target state is generated. When the target state, direction, and time range of two local candidate records are the same, no candidate is generated repeatedly. Each final candidate transition evidence record copies its unique target state from the finite state machine parameter table, specifying the supporting direction, entry threshold, exit threshold, threshold version, and minimum duration corresponding to the two detection positions. Even if one position has not yet met the migration condition in S1, the re-judgment threshold for that position is retained for S4. The transition evidence conflict between the first PPG signal and the second PPG signal includes the single-zone evidence conflict and the direction conflict. Therefore, each candidate transition evidence record corresponds to a defined candidate sleep state transition and a candidate transition window. The candidate sleep state transition is a state transition to be verified, determined according to preset migration conditions but not yet verified and confirmed by differential microload. Multiple records within the same verification event share a single verification trigger, but are processed independently record-by-record in S3 and S4.

[0031] The candidate transition evidence records serve both as the verification trigger input for S2 and as the PPG basic features that S3 must correct and reconstruct, enabling S4 to perform re-judgment using the same threshold version and the same migration conditions at the corresponding detection location. S1 outputs the verification event identifier, the set of candidate transition evidence records, the candidate sleep state transitions and candidate transition windows corresponding to each record, two original transition evidences and their time ranges, and provides this information to S2 to determine whether to initiate short-term verification. If there are no candidate transitions and no evidence conflicts, the current sleep state is maintained and passive data acquisition continues.

[0032] For S2, specifically, after candidate sleep state transitions and candidate transition windows are formed in S1, the controller first performs a verification admission judgment upon receiving a candidate sleep state transition or a conflicting transition evidence. Admission conditions include at least: both local loads have upward and downward adjustment margins, the total contact load has a safety margin, the inertial signal does not characterize large limb movements, both detection locations can continuously detect complete pulses, and the time since the last verification is not less than a preset cooling time. If any condition is not met, the drive is not activated, the current sleep state is maintained, and it is marked as temporarily uncertain; after the movement and contact states stabilize and the cooling time expires, a re-judgment is made for still valid verification events. Thus, during normal monitoring, the differential micro-load component is in a neutral position, only operating briefly when an event is triggered, to reduce the drive duty cycle and disturbance to the wearing area.

[0033] All candidate transformation evidence records under the same verification event identifier trigger only one differential microload verification and share the microload verification data of that verification. After passing the admission judgment, different verification events will each execute at least one complete neutral-first differential-second differential-second differential-first differential-neutral phase sequence. Existing effective local load response relationships under the current contact geometry can only be used to initialize the target change, reduce the search step size, and decrease additional repetition cycles beyond the complete phase sequence; they do not replace the complete verification required for this event, nor do they directly apply previous arbitration results to the current event.

[0034] Before verification begins, the controller reads the first and second actual contact loads as neutral measured loads and determines the target change based on sensor resolution, local safety margin, total load safety margin, and the effective load range of existing relationship records. As a non-limiting example, under the condition that the load sensor range is 0 to 5 Newtons and the resolvable change is no greater than 0.01 Newtons, the single-zone target change can be set to 0.05 to 0.15 Newtons; the target value should also not exceed half of the corresponding local safety margin. The aforementioned values ​​are calibrated using elastic finger molds and force gauges during product assembly and further tightened after the individual resting baseline is established; they cannot replace actual measured load feedback.

[0035] Furthermore, the differential microload verification includes: The neutral phase, the first differential phase, the second differential phase, the second differential phase, the first differential phase, and the neutral phase are executed sequentially. In the first differential phase, the first actual contact load is increased and the second actual contact load is decreased; in the second differential phase, the first actual contact load is decreased and the second actual contact load is increased; each phase continues until a complete pulse is detected. PPG sampling values ​​and actual contact load sampling values ​​are collected in each phase. The PPG sampling values ​​and actual contact load sampling values ​​of each phase are arranged in time to obtain the microload verification data.

[0036] Specifically, the controller establishes phase states in the order of neutral, first differential, second differential, second differential, first differential, and neutral, and only begins accumulating valid samples after the actual load reaches the current phase target and remains within the allowable error band. Each phase covers at least one complete cycle from the pulse start point to the next pulse start point; if the pulse rate is low, the phase duration is extended accordingly, but does not exceed the preset verification timeout. The symmetrical phase sequence ensures that the same differential direction appears once in the first and second halves of the verification, allowing the controller to take the median of two observations in the same direction, thereby mitigating slow physiological drift during verification. Microload verification data is arranged by timestamp, and each record also includes a verification event identifier, phase identifier, detection position identifier, and inertial state, so that S3 can group data according to the actual load increase / decrease direction.

[0037] Furthermore, the differential microload verification specifically includes: Read the local load allowable range and the total load allowable range, and adjust the driving amount of the differential micro-load component according to the feedback of the first actual contact load and the second actual contact load, so that the first actual contact load and the second actual contact load are respectively within the local load allowable range, and the sum of the first actual contact load and the second actual contact load is within the total load allowable range; Inertial signals are collected using inertial sensors. When any of the local contact loads exceeds the local load allowable range, or the sum of the first actual contact load and the second actual contact load exceeds the total load allowable range, or the inertial signal indicates overall slippage, or no complete pulse is detected, the differential micro-load component is controlled to return to a neutral state and stop using the micro-load verification data.

[0038] Specifically, the local load allowable range is the execution value of the preset local safety range in this verification. Its lower limit ensures that the skin-contact pad can still maintain measurable optical contact, and its upper limit is lower than the non-blocking local load upper limit specified in the assembly calibration. The total load allowable range is the execution value of the preset total safety range in this verification, used to limit the impact of the two local adjustments on the overall restraint. The controller can update the drive quantity at 25 Hz. First, it determines the drive direction based on the error between the two local loads and the phase target, and then reduces the drive increment when either local load or total load approaches the boundary. The drive voltage, motor steps, or pumping quantity are only used as execution commands and cannot be used as phase completion criteria. The acceleration and angular velocity output by the inertial sensor are compared with the resting noise upper limit after short-window statistics. If the change exceeds the calibration threshold and is accompanied by a sudden change in the ratio of the two static loads, overall slippage is determined. When any termination condition occurs, the current data is marked as invalid, and the controller returns to the neutral position according to the limited slope. If the feedback does not reach the neutral error band within the timeout, the hardware cuts off the drive power, locks the subsequent verification, and prompts the wearer to remove the device.

[0039] In a preferred embodiment, the differential micro-load assembly includes a piezoelectric bicrystalline beam, a first skin-contact pad, and a second skin-contact pad. The two ends of the piezoelectric bicrystalline beam are mechanically connected to the first skin-contact pad and the second skin-contact pad, respectively. The driving polarity switching of the piezoelectric bicrystalline beam causes the normal displacement directions of the first skin-contact pad and the second skin-contact pad to be interchanged, so as to adjust the first actual contact load and the second actual contact load, respectively.

[0040] Specifically, see Figure 5 The piezoelectric bicrystalline beam is supported in the middle by a wearable shell. The first end is mechanically connected to a first skin-contact pad, and the second end is mechanically connected to a second skin-contact pad. Electrically isolated first and second driving regions are formed on both sides of the central support of the piezoelectric bicrystalline beam. Each driving region includes a piezoelectric layer and partitioned electrodes stacked along the beam's thickness direction. The polarization direction or electrode lead polarity of the second driving region is reversed relative to the first driving region. A differential microload actuator simultaneously applies the same amplitude driving voltage to both driving regions through a bipolar output terminal, causing the two driving regions to undergo bending deformations in opposite directions at the same time. When the first driving polarity is applied, the first end displaces towards the skin while the second end displaces away from the skin. After the driving voltage is reversed, the bending directions of the two driving regions simultaneously reverse, and the actions of the first and second ends are interchanged. When the driving voltage is removed, the elasticity of the beam causes both ends to return to the neutral error zone.

[0041] The first and second skin-contact pads integrate corresponding PPG sensors and load sensors, respectively. Mechanical stops are installed at both ends of the housing along the travel direction, and the drive circuit features hardware current limiting and overvoltage shutdown. During assembly, a force gauge is first used to confirm that the first drive polarity corresponds to load increase at the first end and load decrease at the second end. Then, the reverse polarity is confirmed to correspond to the interchangeable actions, and this polarity-direction relationship is written into the drive configuration. During controller operation, load increase and decrease are still determined in a closed loop based on the measured values ​​of the two load sensors, without estimating the contact load based on the piezoelectric drive voltage. This partitioned reverse drive structure enables a piezoelectric bicrystalline beam to reliably generate reverse normal displacements at both ends, while maintaining a small number of components, short travel, and fast response, making it suitable for the miniaturization of smart rings.

[0042] In another embodiment, the differential micro-load assembly includes an eccentric cam, a rocker arm, a neutral return elastic element, a first skin pad, and a second skin pad. The rocker arm is fulcrumd around the housing of the electronic device. The first skin pad is mechanically connected to a first output end of the rocker arm, and the second skin pad is mechanically connected to a second output end of the rocker arm. The eccentric cam drives the rocker arm through mechanical contact, causing the first output end and the second output end to move along opposite skin normals.

[0043] Specifically, a miniature geared motor drives an eccentric cam to rotate, and the cam profile pushes a double-ended rocker arm with the housing as its fulcrum. When the first output end moves inward, the second output end moves outward, and the motor reverses direction after crossing the neutral angle, thus reversing the action relationship. A torsion spring or leaf spring acts as a neutral return elastic element, pulling the rocker arm back to the neutral position when power is off. The assembly calibration table is only used to convert the motor position into the initial control quantity; the final phase is still confirmed by feedback from two load sensors. Mechanical limits, drive current limits, verification timeouts, and return-to-center fault lockouts are the same as those for the piezoelectric structure. The cam-rocker arm structure and the piezoelectric double crystal beam structure are optional actuators and are not required to be cumulatively installed in the same equipment.

[0044] S2 outputs the micro-load verification data that has passed the validity judgment, as well as the neutral load, the actual load variation range, the contact orientation, and the verification time. If the verification is delayed or terminated, no correctable local response is generated, and all candidate records within the same verification event output uncertain results and maintain the current sleep state; if the verification is valid, the same micro-load verification data along with the candidate conversion evidence record set is handed over to S3.

[0045] For S3, specifically, the controller uses the actual load increase / decrease direction rather than the drive command to phase the microload verification data and reads the candidate transition evidence records for the same verification event one by one. The pulse interval trend in the candidate transition evidence records corresponds to the effective pulse time and pulse interval; the relative change in peripheral vascular tension corresponds to the AC / DC ratio, normalized pulse amplitude, or normalized pulse area; and PPG respiratory modulation corresponds to the selected pulse amplitude or area envelope in the record. For each complete pulse, the controller extracts at least one normalized PPG feature from the first and second PPG signals, including the AC / DC ratio, normalized amplitude, normalized area, or rise time. When the current candidate transition uses relative change in peripheral vascular tension or PPG respiratory modulation as evidence for judgment, the extracted features must include the underlying features used to generate that evidence. The pulse time and pulse interval are stored using the same pulse identifier as the selected normalized PPG features.

[0046] The controller calculates the actual contact load change relative to the adjacent neutral phase for each pulse stability interval. Both the loading and unloading segments should have pairable pulses, and the response direction of the same basic feature at the same detection location should remain consistent within the uncertainty range; otherwise, the local response of the basic feature corresponding to that detection location is invalid. If the basic feature required by the candidate conversion evidence record is missing at any detection location, cannot be paired with the actual load, or does not cover the valid loading and unloading segments, no other irrelevant feature will be used to replace the basic feature; instead, a compensation unavailable flag will be output for that candidate record.

[0047] In some implementations, the step of establishing the dual-zone local load response relationship based on the microload verification data, and using the dual-zone local load response relationship and the local contact load within the candidate conversion window to correct the PPG features within the candidate conversion window, to obtain a dual-zone load normalized PPG feature sequence, includes: Each micro-load phase is composed of data from the micro-load verification data that have the same direction of increase or decrease in local contact load; Normalized PPG features and actual contact load changes are extracted from each micro-load phase, and the local load response relationship of the dual-zone is established based on the normalized PPG features and the actual contact load changes. The static load ratio, the optical DC component, and the electronic device orientation acquired by the inertial sensor are compared between the end of the candidate conversion window and the start time of the differential microload verification. When all differences are within the continuity threshold, a contact geometric continuity determination is obtained. The difference between the first actual contact load within the candidate conversion window and the first actual contact load before the start time of the differential micro-load verification is calculated to obtain the first compensation load. The difference between the second actual contact load within the candidate conversion window and the second actual contact load before the start time of the differential micro-load verification is calculated to obtain the second compensation load. When the result of the contact geometry continuity determination is continuous, and both the first compensation load and the second compensation load are within the load change range of the differential micro-load verification, the PPG features in the candidate conversion window are corrected using the dual-zone local load response relationship, the first compensation load, and the second compensation load to obtain the dual-zone load normalized PPG feature sequence.

[0048] Specifically, the static load ratio can be represented by the ratio of the first actual contact load to the total contact load and the ratio of the second actual contact load to the total contact load; when the total contact load is lower than the reliability calculation lower limit, the evidence for geometric continuity is insufficient. The optical DC component is normalized according to its respective resting baseline, and the orientation of the electronic device is represented by the short-time correction result of the integral of the gravity direction and angular velocity. The continuity threshold can be taken as three times the median absolute deviation of the corresponding change in the individual resting phase, and a minimum threshold allowed by product calibration is set. Only when the three indicators of load distribution, optical coupling, and device orientation are simultaneously continuous is it allowed to use the local response obtained by differential micro-load verification after the candidate conversion window is formed for the candidate conversion window; any discontinuity indicates that ring rotation, removal or contact surface reconstruction may occur, and compensation across this change is prohibited.

[0049] The first and second compensation loads are calculated point-by-point, with their reference being the neutral measured load before the start of verification. The controller records the minimum and maximum load changes actually covered at each detection location during this verification; the compensation load only participates in the correction when it falls within the corresponding coverage area, and can be treated as a boundary response when it equals the boundary, but is not extrapolated when it exceeds the boundary. If there is a contact geometry discontinuity, any compensation load exceeds the boundary, or any local response required by the current candidate is invalid, the determined dual-zone load normalized PPG characteristic sequence of the candidate is not generated, but a compensation unavailable flag is output to S4.

[0050] Furthermore, the first local load response relationship is the median of the ratio of the normalized PPG characteristic change of the first PPG signal to the change of the first actual contact load. The second local load response relationship is the median of the ratio of the normalized PPG characteristic change of the second PPG signal to the change of the second actual contact load; Actual contact load changes whose absolute value is less than the preset load resolution threshold are not included in the median calculation.

[0051] Specifically, to avoid different PPG basic features sharing the same response coefficient, let the detection position number be... The basic feature identifier is The corresponding normalized PPG characteristic change is The corresponding measured load change is Then the detection location For basic features The local load response is determined by the following formula:

[0052] in, , Indicates the first detection position. Indicates the second detection position. It is a basic feature identifier that can correspond to the ratio of AC to DC, normalized pulse amplitude, normalized pulse area, or other normalized PPG basic features specified for the current candidate record. For dimensionless normalized characteristic changes or changes that retain the corresponding characteristic units. The unit is Newton. The units of the basic characteristic m are consistent with the ratio of the units to the Newtons.

[0053] Each response coefficient is associated with and stored as a verification event identifier, detection location, basic feature identifier, unit, effective load range, and version number, and is not universal across different basic features.

[0054] The preset load resolution threshold can be the larger of the assembly calibration resolution and three times the root mean square of the resting load noise, to avoid amplifying noise with small denominators. The controller first removes samples that do not meet the resolution threshold, have incomplete pulses, or have abnormal inertia, and then calculates the median of the ratios of the two detection positions and each basic feature. When the sample covers multiple small load segments, the equivalent local response can also be represented by a piecewise linear lookup table that maintains monotonicity and boundary continuity. When a candidate record specifies multiple basic features, Based on the basic feature identifiers stored in the candidate record, the controller establishes multiple corresponding [records / records]. Each basic feature can only call a response that matches both the detection location and the basic feature identifier; if a matching response is missing, the basic feature is marked as unavailable for compensation.

[0055] Let the candidate conversion window be the first Individual detection locations, basic features The original normalized PPG features are The corresponding actual contact load is The neutral measured load before verification begins is The characteristics after load correction are:

[0056] in, The load normalized PPG basic features of input S4 are used. and The units are all Newtons. and Having the same characteristic units, therefore and The units are consistent. The controller performs the same calculation for each valid sampling point or stroke feature within the candidate transformation window and reconstructs the first and second sequences according to the original timestamps. The pulse time and the pulse interval obtained from the difference between the pulse times are not used. Instead of performing numerical correction, the faulty pulses are removed, and the original timestamps that can be paired are used to recalculate, so as to avoid applying the load response coefficient to the time quantity incorrectly.

[0057] This relationship is only valid within the small load range covered by this validation and within the current contact geometry determined by continuity, and is not used to infer blood pressure, absolute vascular compliance, or clinical sleep stage. Valid relationships, along with load range, orientation summary, optical DC baseline, setup time, and version number, are written to memory; subsequent static load ratios, orientations, or optical DC volumes exceeding the continuity threshold are immediately invalidated.

[0058] To reconstruct the same ternary evidence vector as S1, the controller processes the corrected first and second sequences according to the feature identifiers and basic feature identifiers in each candidate transformed evidence record: It recalculates the pulse interval trend from the original pulse time that has passed quality judgment and completed pairing; it recalculates the relative change in peripheral vascular tension from the AC to DC ratio, normalized amplitude, or normalized area after load correction and identifier matching; and it recalculates the proportion of respiratory band power to the total variable power of the envelope from the pulse amplitude or area envelope selected for the candidate record after load correction, thus obtaining PPG respiratory modulation. The reconstruction result continues to carry the verification event identifier, candidate identifier, target state, source location, support direction, location-corresponding threshold, and threshold version. If any feature involved in the candidate judgment cannot be reconstructed with the same identifier, that feature is marked as missing; if the number of valid reconstructed features at any location is less than the minimum required by the finite state machine, that candidate can only enter the uncertain branch in S4.

[0059] For S4, specifically, the controller reads the candidate transition evidence records one by one and establishes a one-to-one pairing in the first and second sequences corresponding to the record according to the pulse time. The pairing tolerance can be the sum of the upper limit of the synchronization sampling error and the upper limit of the pulse time detection jitter, and this tolerance is limited to no more than half of the median pulse interval between the two locations; when multiple candidates exist within the tolerance, the one with the smallest absolute time difference is selected, and the same pulse cannot be paired repeatedly. Unpaired pulses retain a single-zone identifier and are not forcibly interpolated to be a change in both zones. Direction classification is performed item by item according to the feature identifiers in the candidate transition evidence records, based on the sign of the feature change of each sequence relative to the starting baseline of the candidate window; the measurement uncertainty is calculated from the repeatability characteristics of the resting neutral phase, and the larger of the minimum change caused by sensor resolution and the upper limit of robust dispersion is taken.

[0060] Further, the process of performing pulse pairing and direction classification on the dual-zone load-normalized PPG feature sequence, extracting common physiological components and contact perturbation components, arbitrating the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtaining an arbitration result, and obtaining sleep state monitoring results from the arbitration result includes: The common physiological component is a feature change that is paired with the pulse time, has the same direction, and exceeds the measurement uncertainty. The contact disturbance component is a feature change with opposite direction, a feature change that exists only at the first detection position, or a feature change that exists only at the second detection position. Based on the contact disturbance component, the presence status of the transition evidence for the candidate sleep state transition in the local PPG signal and the dual-zone load normalized PPG feature sequence is compared to obtain the determination of the disappearance of transition evidence. When the common physiological components satisfy the migration conditions for the candidate sleep state transition, the arbitration result is acceptance; otherwise, when the result of the determination of the disappearance of the transition evidence is disappearance, the arbitration result is rejection; otherwise, the arbitration result is indeterminate. The sleep state monitoring results are obtained based on the arbitration result.

[0061] Specifically, after pulse timing pairing, the controller compares the two reconstructed features for each feature identifier in the current candidate transition evidence record. Since the cardiovascular or autonomic nervous system changes corresponding to the same candidate sleep state transition share a common pulse timing reference at both detection locations, and contact load changes primarily affect the optical coupling and local microvascular compression at each detection location, if the feature changes in both regions are in the same direction and their amplitudes are both greater than the corresponding measurement uncertainty, the paired record is written into the common physiological component; if the directions are opposite, it is written into the contact perturbation component; if only one detection location exhibits a change exceeding the uncertainty, it is also written into the contact perturbation component, and the single-region source is retained. If neither region's change exceeds the uncertainty, the feature participating in candidate judgment is missing, or the number of pairs is less than the minimum duration required, the evidence is marked as insufficient. Both the common physiological component and the contact perturbation component store the verification event identifier, candidate identifier, feature identifier, pulse timing, the direction and amplitude of each of the two locations, the source location, and the quality indicator, without pre-combining the two amplitudes into a single value.

[0062] For each of the three pieces of evidence, the controller uses the support direction, entry threshold, exit threshold, and threshold version corresponding to the first and second detection positions to form a position-level support determination. A pair of common evidence pieces is counted as support for the current candidate target only if the following conditions are met simultaneously: the two pulses or analysis windows are paired; both changes are along the support direction of the current candidate; both changes exceed their respective measurement uncertainties; both supporting evidence pieces reach the entry threshold for their corresponding positions; and the above conditions continuously reach the minimum duration window number saved in the candidate record. If the two directions are the same but either one does not reach its entry threshold or minimum duration, the item is only marked as insufficient common change and is not counted as support; if the two directions are opposite or there is only a single-region change, the item is classified as a contact disturbance component. Thus, a pair of position values ​​forms a definite pair of common evidence pieces through position-by-position thresholds, without needing to select an average, maximum, or any single-region value as a substitute.

[0063] The controller re-evaluates the unique target state of the record using the same threshold version from the candidate transition evidence record. Pulse interval trends use the effective pulse interval after two-zone pairing; relative changes in peripheral vascular tone use the corrected matching baseline features; and PPG respiratory modulation uses only the corrected percentage of respiratory band power within the selected pulse amplitude or area envelope in the record. Substitution with other features not recorded in the candidate transition evidence is not permitted. A common physiological component satisfies the candidate's transition criteria when at least two of the three common two-zone evidences satisfy the aforementioned location-level conditions and shortest duration. Thus, the candidate evidence before and after correction has the same feature definitions, baseline features, support direction, location-corresponding thresholds, threshold versions, and time ranges.

[0064] The determination of lost conversion evidence is obtained by comparing the same candidate identifier, the same time range, and the same migration condition before and after correction. If at least two supporting pieces of evidence originally present in the local PPG signal no longer meet the entry threshold or minimum duration at the corresponding position in the dual-zone load-normalized PPG feature sequence, and the corresponding changes have been attributed to the contact perturbation component, then the lost conversion evidence is determined. If at least two pieces of dual-zone common evidence still meet the original migration condition in the corrected common physiological component, then it is not determined to be lost. For each candidate conversion evidence record, arbitration is performed according to a determined priority: if the common physiological component meets the original migration condition, the arbitration result of the record is set to accept; otherwise, it is set to reject when the conversion evidence is lost; otherwise, it is set to uncertain. Geometric discontinuity, compensation exceeding the limit, missing features involved in the judgment, insufficient effective pairing, dual-path still conflicting, or verification aborted all cause the corresponding candidate record to enter the uncertain branch.

[0065] A single verification event generates an event-level result only after all candidate records have been arbitrated record by record. Different verification events undergo the aforementioned arbitration process separately. If an event has only one candidate record, the record's acceptance, rejection, or indeterminate result is directly adopted. If an event contains multiple candidate records, and exactly one candidate record is accepted while all the remaining candidate records are rejected, the sleep state is updated to the unique target state of the accepted record. If no candidate record is accepted and all candidate records are rejected, the original sleep state is maintained and a contact artifact is recorded. If any candidate record is indeterminate, or two or more distinct target states are accepted simultaneously, the event-level result is indeterminate and the original sleep state is maintained. This rule ensures that dual-zone directional conflicts do not yield different results due to different processing orders of candidate records.

[0066] The sleep state monitoring results include at least the current sleep state, event-level arbitration result, accepted candidate identifier or null value, candidate transition time, contact artifact flag, and uncertainty interval. Acceptance causes the finite state machine to transition to the only accepted candidate state; rejection of all results deletes the candidate transition but not the original sample; uncertain results retain the original state, and full verification is performed on new, still valid verification events after contact stabilization and the cooling time has elapsed. After verification, the differential microload component returns to a neutral state.

[0067] If the event is accepted or rejected entirely, and the static load ratio, optical DC component, and equipment orientation continue to meet the contact geometric continuity condition, the controller extends the validity period of the local load response relationship and increases the cooling time for repeated triggers within the same target and time neighborhood. This only suppresses jittery repeated triggers of the same event and does not skip different verification events formed after the cooling time. If the event-level result is uncertain, the validity period of the relationship is not extended, nor is the cooling time increased accordingly. When any continuity flag exceeds the threshold, the controller immediately cancels the old relationship, restores the default target change amount, and the default cooling time. Thus, relationship reuse reduces unnecessary verification frequency by merging multiple candidates for the same event, initializing the verification amplitude, and suppressing repeated triggers of the same event, while retaining at least one complete verification for each different verification event. Wireless communication can upload results and quality flags; if communication fails, they are cached locally and retransmitted after the connection is restored.

[0068] This embodiment also provides an electronic device for monitoring sleep states.

[0069] In some embodiments, the device includes a wearable housing; a first measurement component comprising a first skin-contact pad, a first PPG sensor, and a first load sensor, disposed at a first detection position; a second measurement component comprising a second skin-contact pad, a second PPG sensor, and a second load sensor, disposed at a second detection position; the first measurement component and the second measurement component being spaced apart on the same skin-contact side of the wearable housing; a passive support pad disposed on opposite sides of the first measurement component and the second measurement component; a differential micro-load component mechanically connected to the first skin-contact pad and the second skin-contact pad respectively, for moving the first skin-contact pad and the second skin-contact pad along opposite skin normals; an inertial sensor; a synchronous sampling front end electrically connected to the first PPG sensor, the second PPG sensor, the first load sensor, the second load sensor, and the inertial sensor respectively; and a controller electrically connected to the synchronous sampling front end and the differential micro-load component respectively, for executing the aforementioned method.

[0070] Specifically, see Figure 4 The wearable housing defines the boundaries of a single electronic device. A first PPG sensor and a first load sensor form the optical and load inputs for a first detection position via a synchronous sampling front end. A second PPG sensor and a second load sensor form the optical and load inputs for a second detection position via the same synchronous sampling front end. An inertial sensor is also connected to this front end or latched by a unified clock. The synchronous sampling front end may include an LED driver circuit, a transimpedance amplifier circuit, an anti-aliasing filter circuit, a multiplexer analog-to-digital converter, and a sampling timing circuit. Its output is transmitted to the controller via an internal bus. The passive support pad is a mechanical component connected to the housing and does not require electrical connection; it forms a stable contact fulcrum with the two measuring components on the same side.

[0071] The memory is connected to the controller via an internal bus and stores the program instructions that enable the controller to execute S1 to S4, the finite state machine parameter table, the verification event record, and the candidate transition evidence record. During S1 execution, the controller reads the dual-zone PPG, dual-zone load, and inertial frame with a unified timestamp from the synchronous sampling front end, establishes a candidate transition window, generates one or more single-target candidate records according to defined rules, and saves the feature identifier and threshold version. During S2 execution, the controller sends only one set of phase commands for a single verification event and uses the returned dual-zone measured load and inertial state closed-loop constraint to limit the local load and total load. During S3 execution, the controller reads the candidate record set, each candidate window, and shared verification data from the memory, establishes the dual-zone local load response relationship with an effective domain according to the detection position and basic feature identifier, reconstructs the defined ternary evidence vector for each candidate, and outputs the dual-zone load normalized PPG feature sequence. During S4 execution, the controller completes pulse pairing, position-by-position threshold judgment, candidate-by-candidate arbitration, and event-level unique summarization. The differential microload driver is electrically connected to the controller and also electrically connects to or drives the differential microload assembly, which is mechanically connected to the two skin-contact pads. The wireless communication unit is an optional result interface and does not participate in the load safety closed loop.

[0072] use Figure 5 When using the piezoelectric bicrystalline beam, the bipolar output terminals of the differential microload actuator are connected to the partition electrodes of the first and second driving regions, respectively, and simultaneously reverse according to the reverse lead relationship saved in the assembly calibration. When using a cam-rocker arm or microfluidic structure, the differential microload actuator outputs motor commutation commands or micropump and directional valve control signals, respectively. All three execution routes use the same phase state, measured load feedback, and safety stop interface, thus not changing the data link for the controller to execute S2 to S4.

[0073] During the circuit assembly phase, the zero point, range, resolution, differential travel, total load fluctuation, mechanical stop, and homing error band of the two load sensors can be calibrated using flexible fingerprints and traceable force gauges. Once worn, the controller establishes a personal optical baseline, load noise, and azimuth baseline within a resting window with minimal movement and a stable pulse. In the event of a single PPG or load channel failure, the device retains coarse-grained awake or low-activity recordings but does not confirm fine-grained sleep state transitions. In the event of low battery voltage, drive overcurrent, or homing failure, the hardware shuts down the drive and maintains passive acquisition, outputting a removal prompt if necessary. Thus, program processing, drive actions, and mechanical limits together form a safety chain that allows for independent operation.

[0074] As another executable carrier, the wearable shell can be constructed as a smart wristband, with two skin-contact pads spaced apart along the width of the band, forming an equal-volume and interconnected first and second flexible liquid chambers beneath them. When a bidirectional micropump or reversing valve transfers a quantitative amount of liquid from the first flexible liquid chamber to the second flexible liquid chamber, the first detection position is unloaded while the second detection position is loaded; the actions are reversed during the reverse transfer. A fixed total liquid volume is used to reduce overall constraint changes, but the phase completion, safety boundary, and local response calculations are still based on the measured feedback from the two load sensors. This microfluidic structure, piezoelectric bicrystalline beam structure, and cam-rocker arm structure can be selected to implement differential micro-loads without changing the candidate triggering of S1, the local and total load closed loop of S2, the current geometrically bounded correction of S3, or the dual-zone direction arbitration of S4.

[0075] From the overall technical mechanism perspective, low-power passive acquisition first retains candidate windows for potential sleep state changes, and then fixes the ternary evidence upon which each candidate is based with candidate transition evidence records containing feature identifiers, threshold versions, and unique target states. One or more candidate records from the same time period are grouped into the same verification event, and then a short-term, reversed spatial perturbation constrained by safety boundaries is applied when candidate transitions or dual-zone evidence conflicts. Local contact changes exhibit a response related to the differential direction at the two detection locations, while common cardiovascular or autonomic changes have a pairable pulse time reference. Therefore, the measured load response established according to the detection location and basic features under the current wearing geometry can subtract the local load contribution from the corresponding basic features of the candidate window, and form a common physiological component by utilizing the temporal consistency, directional consistency, and position-by-position threshold of the corrected dual-zone changes. The corrected baseline features are recalculated as ternary evidence with the same definition, direction, and threshold as the candidate stages. This allows the finite state machine to accept, reject, or postpone state transitions on a candidate-by-candidate basis. Conflicts between opposing objectives are resolved through unique event-level aggregation, avoiding data gaps created by simply deleting pressure anomaly windows. It also prevents misinterpreting local PPG changes caused by hand pressure, positional shifts, or nocturnal edema as definitive sleep or wakefulness. Event triggering, localized increases and decreases, and total load limiting further eliminate the need for continuous pressure on the wearing area during verification and reduce the drive duty cycle.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring sleep states, characterized in that, The local PPG signal includes a first PPG signal and a second PPG signal; the local contact load includes a first actual contact load and a second actual contact load; the detection position includes a first detection position and a second detection position; and the dual-zone local load response relationship includes a first local load response relationship and a second local load response relationship. The first PPG signal and the first actual contact load originate from the first detection position; the second PPG signal and the second actual contact load originate from the second detection position; the first local load response relationship belongs to the first detection position; and the second local load response relationship belongs to the second detection position. The sum of the first actual contact load and the second actual contact load is the total contact load; the method includes: S1. Acquire the local PPG signal and the local contact load, and obtain candidate sleep state transitions and candidate transition windows based on the local PPG signal; S2. In response to the candidate sleep state transition or the conflict of transition evidence between the first PPG signal and the second PPG signal, control the differential microload component to perform differential microload verification. The differential microload verification adjusts the local contact load by increasing and decreasing it in reverse according to the feedback of the local contact load and the total contact load, limiting the local contact load to a preset local safety range and the total contact load to a preset total safety range. The local PPG signal and the local contact load during the adjustment period are collected to obtain microload verification data. S3. Establish the dual-zone local load response relationship based on the microload verification data. Use the dual-zone local load response relationship and the local contact load in the candidate conversion window to correct the PPG features in the candidate conversion window and obtain the dual-zone load normalized PPG feature sequence. S4. Perform pulse pairing and direction classification on the dual-zone load normalized PPG feature sequence, extract common physiological components and contact perturbation components, arbitrate the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtain the arbitration result, and obtain the sleep state monitoring result from the arbitration result.

2. The method according to claim 1, characterized in that, The step of obtaining candidate sleep state transitions and candidate transition windows based on the local PPG signal includes: Pulse interval trends, relative changes in peripheral vascular tension, and PPG respiratory modulation were extracted from the first PPG signal and the second PPG signal. Using the pulse interval trend, the relative change in peripheral vascular tension, and the PPG respiratory modulation as inputs, the candidate sleep state transitions are obtained using a finite state machine with an entry threshold, an exit threshold, and a minimum duration, and the continuous sampling intervals that meet the transition conditions are determined as the candidate transition windows.

3. The method according to claim 1, characterized in that, The differential microload verification includes: The neutral phase, the first differential phase, the second differential phase, the second differential phase, the first differential phase, and the neutral phase are executed sequentially. In the first differential phase, the first actual contact load is increased and the second actual contact load is decreased; in the second differential phase, the first actual contact load is decreased and the second actual contact load is increased; each phase continues until a complete pulse is detected. PPG sampling values ​​and actual contact load sampling values ​​are collected in each phase. The PPG sampling values ​​and actual contact load sampling values ​​of each phase are arranged in time to obtain the microload verification data.

4. The method according to claim 3, characterized in that, The differential microload verification specifically includes: Read the local load allowable range and the total load allowable range, and adjust the driving amount of the differential micro-load component according to the feedback of the first actual contact load and the second actual contact load, so that the first actual contact load and the second actual contact load are respectively within the local load allowable range, and the sum of the first actual contact load and the second actual contact load is within the total load allowable range; Inertial signals are collected using inertial sensors. When any of the local contact loads exceeds the local load allowable range, or the sum of the first actual contact load and the second actual contact load exceeds the total load allowable range, or the inertial signal indicates overall slippage, or no complete pulse is detected, the differential micro-load component is controlled to return to a neutral state and stop using the micro-load verification data.

5. The method according to claim 1, characterized in that, The differential micro-load assembly includes a piezoelectric bicrystalline beam, a first skin-contact pad, and a second skin-contact pad. The two ends of the piezoelectric bicrystalline beam are mechanically connected to the first skin-contact pad and the second skin-contact pad, respectively. The driving polarity switching of the piezoelectric bicrystalline beam causes the normal displacement directions of the first skin-contact pad and the second skin-contact pad to be interchanged, so as to adjust the first actual contact load and the second actual contact load, respectively.

6. The method according to claim 1, characterized in that, The differential micro-load assembly includes an eccentric cam, a rocker arm, a neutral return elastic element, a first skin pad, and a second skin pad. The rocker arm is fulcrumd with the housing of the electronic device. The first skin pad is mechanically connected to the first output end of the rocker arm, and the second skin pad is mechanically connected to the second output end of the rocker arm. The eccentric cam drives the rocker arm through mechanical contact, causing the first output end and the second output end to move along opposite skin normals.

7. The method according to claim 1, characterized in that, The step of establishing the dual-zone local load response relationship based on the microload verification data, and using the dual-zone local load response relationship and the local contact load within the candidate conversion window to correct the PPG features within the candidate conversion window, yields a dual-zone load normalized PPG feature sequence, including: Each micro-load phase is composed of data from the micro-load verification data that have the same direction of increase or decrease in local contact load; Normalized PPG features and actual contact load changes are extracted from each micro-load phase, and the local load response relationship of the dual-zone is established based on the normalized PPG features and the actual contact load changes. The static load ratio, the optical DC component, and the electronic device orientation acquired by the inertial sensor are compared between the end of the candidate conversion window and the start time of the differential microload verification. When all differences are within the continuity threshold, a contact geometric continuity determination is obtained. The difference between the first actual contact load within the candidate conversion window and the first actual contact load before the start time of the differential micro-load verification is calculated to obtain the first compensation load. The difference between the second actual contact load within the candidate conversion window and the second actual contact load before the start time of the differential micro-load verification is calculated to obtain the second compensation load. When the result of the contact geometry continuity determination is continuous, and both the first compensation load and the second compensation load are within the load change range of the differential micro-load verification, the PPG features in the candidate conversion window are corrected using the dual-zone local load response relationship, the first compensation load, and the second compensation load to obtain the dual-zone load normalized PPG feature sequence.

8. The method according to claim 7, characterized in that, The first local load response relationship is the median of the ratio of the normalized PPG characteristic change of the first PPG signal to the change of the first actual contact load. The second local load response relationship is the median of the ratio of the normalized PPG characteristic change of the second PPG signal to the change of the second actual contact load; Actual contact load changes whose absolute value is less than the preset load resolution threshold are not included in the median calculation.

9. The method according to claim 1, characterized in that, The process involves performing pulse pairing and direction classification on the dual-zone load-normalized PPG feature sequence, extracting common physiological components and contact perturbation components, arbitrating the candidate sleep state transitions based on the common physiological components and the contact perturbation components, obtaining an arbitration result, and obtaining sleep state monitoring results from the arbitration result, including: The common physiological component is a feature change that is paired with the pulse time, has the same direction, and exceeds the measurement uncertainty. The contact disturbance component is a feature change with opposite direction, a feature change that exists only at the first detection position, or a feature change that exists only at the second detection position. Based on the contact disturbance component, the presence status of the transition evidence for the candidate sleep state transition in the local PPG signal and the dual-zone load normalized PPG feature sequence is compared to obtain the determination of the disappearance of transition evidence. When the common physiological components satisfy the migration conditions for the candidate sleep state transition, the arbitration result is acceptance; otherwise, when the result of the determination of the disappearance of the transition evidence is disappearance, the arbitration result is rejection; otherwise, the arbitration result is indeterminate. The sleep state monitoring results are obtained based on the arbitration result.

10. An electronic device for monitoring sleep states, characterized in that, The device includes a wearable housing; a first measurement component comprising a first skin-contact pad, a first PPG sensor, and a first load sensor, disposed at a first detection position; a second measurement component comprising a second skin-contact pad, a second PPG sensor, and a second load sensor, disposed at a second detection position; the first measurement component and the second measurement component are spaced apart and disposed on the same skin-contact side of the wearable housing; a passive support pad disposed on opposite sides of the first measurement component and the second measurement component; a differential micro-load component mechanically connected to the first skin-contact pad and the second skin-contact pad respectively, for moving the first skin-contact pad and the second skin-contact pad along opposite skin normals; an inertial sensor; a synchronous sampling front end electrically connected to the first PPG sensor, the second PPG sensor, the first load sensor, the second load sensor, and the inertial sensor respectively; and a controller electrically connected to the synchronous sampling front end and the differential micro-load component respectively, for executing the method of any one of claims 1 to 9.

Images