Portable blood oxygen signal OSA intelligent detection method

Abstract

The invention relates to the technical field of medical monitoring, and discloses a portable blood oxygen signal OSA intelligent detection method which comprises the following steps: acquiring a pulse wave signal through a blood oxygen sensor, generating an amplitude difference sequence and a period difference sequence in parallel, analyzing rhythm synchronism of the amplitude difference sequence and the period difference sequence, and judging a respiratory effort event when a real-time relation index deviates from a reference value. By capturing a pulse wave time sequence-amplitude decoupling phenomenon, direct perception of a specific mode process of an invalid respiratory effort signal is realized, a traditional hysteretic monitoring mode depending on blood oxygen decline is avoided, and meanwhile, the system has the internal capability of distinguishing physiological fluctuation and specific mode struggling of the signal, and is convenient to use. And a brand new solution is provided for early screening of light and moderate OSA.

Description

Technical Field

[0001] The present invention relates to a portable blood oxygen signal OSA intelligent detection method, belonging to the technical field of medical monitoring. Background Art

[0002] In the field of portable remote monitoring of sleep breathing disorders, existing technologies mainly rely on physiological result signals such as decreased blood oxygen saturation and heart rate variability as the basis for determining obstructive sleep apnea (OSA). Whether through polysomnography or simplified portable devices, their common limitation is that they focus on the delayed physiological compensation phenomenon after airway obstruction, such as decreased blood oxygen or pulse rate fluctuations, but cannot capture the ineffective breathing effort process that occurs in the early stage of obstruction. When facing patients with mild to moderate OSA, such solutions become ineffective during the nighttime struggle stage when the patients have not yet shown significant abnormal indicators, resulting in the missed diagnosis of a large number of hidden cases.

[0003] Further analysis shows that existing technical methods have fundamental constraints: their signal processing logic passively relies on physiological results rather than the specific pattern process of the signal itself. For example, conventional solutions eliminate respiratory pulse wave modulation through filtering, but ignore the real-time airway dynamics information contained in the modulated signal. Even if a multi-sensor fusion strategy is introduced to improve accuracy, it is still difficult to deploy on edge devices due to its reliance on complex algorithms and high computing power, and it is impossible to distinguish between physiological fluctuations such as changes in body position and the specific pattern of breathing struggles of the real signal.

[0004] Specifically, existing technologies face three core bottlenecks: 1. They can only identify obstructions that have already occurred, but cannot perceive ongoing respiratory effort; 2. They have difficulty distinguishing the essential difference between physiological interference and the specific pattern of struggle in the signal; and 3. Complex models are difficult to adapt to the hardware constraints of ultra-large-scale home screening scenarios. Therefore, the technical problem to be solved by this invention is to provide a method that can sensitively reflect the specific pattern of respiratory rhythm changes in pulse wave signals. Summary of the Invention

[0005] The present invention provides a portable blood oxygen signal OSA intelligent detection method, the main purpose of which is to solve the problem of how to directly capture the specific pattern process of the respiratory effort signal through a single blood oxygen signal and achieve high robustness classification and recognition of edge devices.

[0006] To achieve the above objectives, the present invention provides a portable blood oxygen signal OSA intelligent detection method, the method comprising the following steps: Step a, continuously collecting photoplethysmography signals and detecting the pulse wave peak value of the photoplethysmography signals in real time to obtain a pulse wave peak sequence; Step b, based on the pulse wave peak sequence, generating in parallel an amplitude difference sequence reflecting the change of the pulse wave amplitude and a period difference sequence reflecting the change of the pulse cycle interval; Step c: analyzing the rhythmic synchronization of the amplitude difference sequence and the period difference sequence within a preset time window. The rhythmic synchronization is obtained by calculating the sign consistency ratio of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence. The sign consistency ratio represents the degree of sign consistency of the two sequences. Step d, comparing the symbol consistency ratio obtained in real time with a reference symbol consistency ratio, where the reference symbol consistency ratio is determined under a stable breathing state at the initial stage of user monitoring; Step e: when the real-time symbol consistency ratio is continuously lower than the first preset threshold value of the reference symbol consistency ratio, the current signal segment is determined to be a rhythm desynchronization state signal segment.

[0007] Preferably, in step c, the symbol consistency ratio is calculated as follows: within a preset time window, count the number of times the sign of the product of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence is positive, and take the ratio of the number to the total number of heartbeats in the time window as the symbol consistency ratio.

[0008] Preferably, the reference symbol consistency ratio is adaptively established by calculating a moving average of the symbol consistency ratio in a stable breathing state at the initial stage of user monitoring.

[0009] Preferably, the method further includes the following steps: when the cumulative duration or occurrence frequency of the determined rhythm desynchronization state signal segment reaches or exceeds a preset reference threshold, outputting a high-risk level warning.

[0010] Preferably, after step e determines that a rhythm desynchronization state signal segment occurs, the following steps are also included: calculating the second-order differential value of the amplitude difference sequence and the second-order differential value of the period difference sequence at the time point when the rhythm desynchronization state signal segment occurs; if any of the absolute values ​​of the second-order differential values ​​of the amplitude difference sequence or the absolute values ​​of the second-order differential values ​​of the period difference sequence is greater than a preset morphological identification threshold, the event is marked as a high-frequency interference segment, and its weight is reduced or excluded when calculating the risk index.

[0011] Preferably, the preset morphological discrimination threshold is set according to historical physiological data, and is used to distinguish between relatively gentle changes caused by respiratory effort and instantaneous drastic changes caused by high-frequency signal disturbances of non-respiratory sources.

[0012] Preferably, the method also includes the following steps: continuously monitoring the user's body movement through an acceleration sensor; and when the body movement amplitude is detected to be greater than a preset body movement threshold, suspending the judgment of step e to avoid interference of body movement artifacts in the judgment of the rhythm desynchronization state signal segment.

[0013] Preferably, the method further includes the following steps: in the time period when the rhythm desynchronization state signal segment is not determined in step e, continuously calculating the modulation energy of the amplitude difference sequence and the modulation energy of the period difference sequence, wherein the modulation energy is obtained by calculating the variance of the difference sequence in the time period; when the modulation energy of the amplitude difference sequence is less than and the modulation energy of the periodic difference sequence At the same time, when the energy is continuously lower than the respective energy baseline lower limit threshold within a preset duration, a specific signal segment representing the silence of the signal modulation energy is identified.

[0014] Preferably, the energy baseline lower limit threshold is set to the lower limit ratio value of the corresponding modulation energy in the user's normal stable sleep state, and the range of the lower limit ratio value is 0.1 to 0.3.

[0015] Preferably, the modulation energy baseline in a normal stable sleep state is adaptively established by performing a moving average on the variance of the amplitude difference sequence and the period difference sequence in a stable breathing state at the initial stage of user monitoring.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. By analyzing the rhythmic coupling relationship between pulse wave amplitude and period, the present invention proposes a new signal feature (such as the symbol consistency ratio). Compared with the traditional blood oxygen saturation index, this feature can reflect the changes in respiratory rhythm earlier and more sensitively. This method provides a new technical means for the dynamic monitoring of respiratory status and can effectively distinguish signal interference caused by body movement, etc., and the specific pattern of the signal. At the same time, based on the symbol consistency analysis of the pulse wave differential sequence, the system naturally distinguishes physiological fluctuations (such as turning over, sleep stage transitions) from the specific pattern of respiratory struggles in the signal. When the body motion sensor pause logic is synchronously introduced, the judgment process is further dynamically controlled through the motion amplitude threshold, so that body motion noise no longer triggers false judgments. This synergy between timing rhythm analysis and physical movement monitoring establishes a double anti-interference barrier at the signal essence level, and still maintains recognition reliability in the multi-disturbance environment at home.

[0017] 2. Reusing the same set of pulse wave feature data streams, the system activates the central apnea determination logic by monitoring the abnormal attenuation of differential sequence modulation energy while identifying obstructive events. This specific patterned interpretation of silent signals enables a single blood oxygen sensor to distinguish between obstructive and central apnea for the first time, improving the differentiation of different signal patterns and providing professionals with richer technical information for subsequent analysis.

[0018] 3. The rhythm synchronization baseline established during the user's resting phase is adaptively updated through moving average during continuous monitoring. When combined with the second-order differential morphological analysis of the pulse wave differential sequence, the system can autonomously identify transient artifacts caused by high-frequency signal disturbances from non-respiratory sources, thus preventing ECG interference from contaminating the respiratory event library. This integrated timing management mechanism, which integrates calibration, monitoring, and verification, significantly reduces the device's dependence on idealized usage conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a timing diagram of body motion amplitude interference recognition and respiratory analysis control in the present invention; Figure 2 This is a time series analysis diagram of rhythm and modulation activity of the present invention, where Figure 2 (a) is a schematic diagram showing the change of rhythm synchronization ratio over time. Figure 2 (b) Schematic diagram of modulation activity in different sleep stages; Figure 3 This is a respiratory event determination flow chart of the present invention; Figure 4 This is a 24-hour monitoring curve chart of the symbol consistency ratio of the present invention; Figure 5 This is a comparative analysis diagram of the differential sequence modulation energy of the present invention; Figure 6 This is a timing diagram of body motion interference and system status in the present invention.

[0020] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0021] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0022] The present invention provides a portable blood oxygen signal OSA intelligent detection method, which includes the following steps: Step a, continuously collecting photoplethysmography signals and detecting the pulse wave peak value of the photoplethysmography signals in real time to obtain a pulse wave peak sequence; Step b, based on the pulse wave peak sequence, generating in parallel an amplitude difference sequence reflecting the change of the pulse wave amplitude and a period difference sequence reflecting the change of the pulse cycle interval; Step c: analyzing the rhythmic synchronization of the amplitude difference sequence and the period difference sequence within a preset time window. The rhythmic synchronization is obtained by calculating the sign consistency ratio of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence. The sign consistency ratio represents the degree of sign consistency of the two sequences. Step d, comparing the symbol consistency ratio obtained in real time with a reference symbol consistency ratio, where the reference symbol consistency ratio is determined under a stable breathing state at the initial stage of user monitoring; Step e: when the real-time symbol consistency ratio is continuously lower than the first preset threshold value of the reference symbol consistency ratio, the current signal segment is determined to be a rhythm desynchronization state signal segment.

[0023] Preferably, in step c, the symbol consistency ratio is calculated as follows: within a preset time window, count the number of times the sign of the product of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence is positive, and take the ratio of the number to the total number of heartbeats in the time window as the symbol consistency ratio.

[0024] Preferably, the reference symbol consistency ratio is adaptively established by calculating a moving average of the symbol consistency ratio in a stable breathing state at the initial stage of user monitoring.

[0025] Preferably, the method further includes the following steps: when the cumulative duration or occurrence frequency of the determined rhythm desynchronization state signal segment reaches or exceeds a preset reference threshold, outputting a high-risk level warning.

[0026] Preferably, after step e determines that a rhythm desynchronization state signal segment occurs, the following steps are also included: calculating the second-order differential value of the amplitude difference sequence and the second-order differential value of the period difference sequence at the time point when the rhythm desynchronization state signal segment occurs; if any of the absolute values ​​of the second-order differential values ​​of the amplitude difference sequence or the absolute values ​​of the second-order differential values ​​of the period difference sequence is greater than a preset morphological identification threshold, the event is marked as a high-frequency interference segment, and its weight is reduced or excluded when calculating the risk index.

[0027] Preferably, the preset morphological discrimination threshold is set according to historical physiological data, and is used to distinguish between relatively gentle changes caused by respiratory effort and instantaneous drastic changes caused by high-frequency signal disturbances of non-respiratory sources.

[0028] Preferably, the method also includes the following steps: continuously monitoring the user's body movement through an acceleration sensor; and when the body movement amplitude is detected to be greater than a preset body movement threshold, suspending the judgment of step e to avoid interference of body movement artifacts in the judgment of the rhythm desynchronization state signal segment.

[0029] Preferably, the method further includes the following steps: in the time period when the rhythm desynchronization state signal segment is not determined in step e, continuously calculating the modulation energy of the amplitude difference sequence and the modulation energy of the period difference sequence, wherein the modulation energy is obtained by calculating the variance of the difference sequence in the time period; when the modulation energy of the amplitude difference sequence is less than and the modulation energy of the periodic difference sequence At the same time, when the energy is continuously lower than the respective energy baseline lower limit threshold within a preset duration, a specific signal segment representing the silence of the signal modulation energy is identified.

[0030] Preferably, the energy baseline lower limit threshold is set to the lower limit ratio value of the corresponding modulation energy in the user's normal stable sleep state, and the range of the lower limit ratio value is 0.1 to 0.3.

[0031] Preferably, the modulation energy baseline in a normal stable sleep state is adaptively established by taking a moving average of the variances of the amplitude difference sequence and the period difference sequence under a stable breathing state at the initial stage of user monitoring; at the same time, in the present invention, the setting of the reference symbol consistency ratio is based on the stable breathing state of the user in the initial monitoring stage, and is adaptively established by calculating the moving average of the symbol consistency ratio at this time. This method can dynamically adjust the reference value to adapt to individual physiological differences. In specific applications, when the user is in a resting state at the initial stage of monitoring, a preliminary baseline value is calculated through stable respiratory status data. Subsequently, the system will adaptively update the user's long-term monitoring data. The update cycle of the baseline value is closely related to the stability of the monitoring data. When the system detects a significant change in the respiratory state, the baseline value will be revised according to the new data to ensure that the respiratory characteristics of each user are fully considered; for the preset thresholds, the first consideration is the adaptability to the individual differences of different users. The setting of these thresholds is not only based on the percentage change of the baseline symbol consistency ratio, but also needs to take into account the user's individual physiological characteristics. The selection range of the preset threshold is usually between 20% and 30% of the baseline value. The specific range is optimized according to actual tests and the user's physiological responses. The threshold range is based on the engineering trade-off between normal fluctuations in respiratory rhythm changes and possible specific pattern changes of the signal to ensure that the system can effectively identify respiratory abnormalities in patients with mild to moderate OSA while avoiding misjudging normal physiological fluctuations. The reasonable setting of the threshold is to achieve the best balance between sensitivity and specificity to ensure the accuracy of recognition. In terms of event recognition, the present invention judges the ineffective respiratory effort event and the rhythm desynchronization state signal segment by analyzing the rhythm synchronization of the amplitude difference sequence and the period difference sequence. The ineffective respiratory effort event usually refers to the situation where the patient's respiratory effort fails to effectively lead to the improvement of airflow when the airway is blocked. This type of event is characterized by the occurrence of rhythm decoupling, which is reflected by a significant decrease in the sign consistency ratio. This phenomenon can be detected in the early stage before the blood oxygen saturation decreases, and has become an important method for early screening of mild to moderate OSA. The rhythm desynchronization state signal segment refers to the respiratory effort event caused by complete or partial airway obstruction, which is characterized by a continuous decrease in the symbol consistency ratio when monitoring the blood oxygen signal, accompanied by an increase in physiological interference. Therefore, by comparing the rhythm synchronization ratio with the baseline value, the system can effectively distinguish between the two events. To ensure the accuracy of this process, the system will also combine the second-order difference calculation and morphological identification mechanism to identify different types of artifacts. For example, when the second-order difference value of the signal exceeds the preset morphological identification threshold, the event will be marked as a cardiac artifact and excluded from the final judgment.The morphological discrimination threshold is set based on the fluctuation patterns of historical physiological data, aiming to distinguish between gentle fluctuations caused by respiratory effort and sharp fluctuations caused by high-frequency signal disturbances from non-respiratory sources. Furthermore, in the present invention, to prevent motion artifacts from interfering with the determination of rhythm desynchronization signal segments, the system monitors the user's body motion in real time via an accelerometer and determines whether to suspend event determination based on a preset motion amplitude threshold. When the motion amplitude exceeds the preset motion threshold, the system suspends respiratory event determination for the current data segment. The motion threshold is selected based on the typical range of motion amplitude variation at the device's wear position. Specifically, the motion amplitude threshold is typically set at approximately 0.5G to ensure that artifacts are not misidentified within the normal range of motion. The motion monitoring setting also takes into account adaptability to different usage environments. For example, motion amplitude varies across different sleep stages, sleeping postures, and living environments. Therefore, the system adjusts the motion threshold based on the user's long-term monitoring data to minimize the impact of motion artifacts in varying environments. These are all extended implementations known to those skilled in the art.

[0032] Example 1: This example proposes a portable blood oxygen signal OSA intelligent detection method. Based on the photoelectric volumetric pulse wave signal collected by a single blood oxygen sensor, the method extracts the pulse wave peak sequence and constructs its corresponding amplitude difference sequence and period difference sequence. The degree of synchronization of these two types of difference sequences at the rhythm level is further calculated, and the rhythm desynchronization state signal segment is identified accordingly. On this basis, artifacts are eliminated in combination with morphological features, and when specific accumulation conditions are met, the specific signal segment representing the silence of signal modulation energy is determined. Specifically, in the signal acquisition stage, a conventional photoelectric volumetric sensor is used to continuously sample the user's pulse wave signal. It is usually worn on the fingertips or earlobes for photoelectric detection to obtain a pulse waveform reflecting cardiac activity. The system extracts the peak of the pulse waveform in real time to construct a pulse wave peak sequence. This peak sequence provides basic data for subsequent differential processing. Its stability directly affects the accuracy of the entire determination process. In actual implementation, a multi-scale smoothing and pseudo-peak correction strategy can be adopted. To improve the robustness of peak extraction under low signal-to-noise ratio conditions; in addition, during the construction of the pulse wave peak sequence, to ensure the stability and accuracy of peak extraction, the amplitude threshold setting is based on the average pulse wave peak-to-valley difference of the input signal in the static segment, and is typically set to 50% of the average value to eliminate pseudo-peak responses caused by transient disturbances or noise. In conjunction with this threshold strategy, the system introduces a pseudo-peak correction mechanism, which specifically includes: first, after the peak candidate point is initially detected, a time window is set according to the typical duration of adjacent cardiac cycles to determine whether the candidate point constitutes a local maximum center; if there are multiple adjacent candidate points, the one with the largest amplitude within the local time window is retained, and the rest are eliminated; second, if the distance between two adjacent peaks is significantly lower than the lower limit of the average cardiac cycle, the system will trigger the adjacent peak re-comparison and replacement process to correct the pseudo-dense peak structure induced by high-frequency interference, ensuring that the final pulse wave peak sequence has good time alignment characteristics and amplitude representativeness.Based on the aforementioned peak sequence, the system constructs two differential sequences: an amplitude differential sequence, representing the amplitude variations between consecutive pulse wave peaks, whose elements consist of the differences between adjacent peaks; and a period differential sequence, reflecting the variations in the time intervals between peaks, whose elements consist of the variations in the intervals between consecutive heartbeats. These two differential sequences reveal the local volatility of the signal in both the amplitude and time domains, respectively, and serve as key characteristic dimensions for identifying the specific pattern of respiratory effort associated with the signal. The system sets a sliding time window to calculate the rhythm synchronization index, namely the sign consistency ratio, between the two differential sequences during real-time monitoring. This ratio is defined as the number of times within the time window that the product of the corresponding elements of the amplitude differential sequence and the period differential sequence is positive, and then the ratio of this number to the total number of heartbeats within the time window. The resulting sign consistency ratio is used to quantify the degree of consistency between the two differential sequences in the direction of sign fluctuation, thereby reflecting the rhythmic coupling relationship between the respiratory modulation signal and the pulse wave.

[0033] Under normal breathing conditions, there is relatively stable rhythmic synchronization between amplitude difference and period difference, and the sign consistency ratio is usually maintained at a high level. However, when a rhythm desynchronization signal segment occurs, airway obstruction causes abnormal fluctuations in negative pressure in the lungs, resulting in rhythmic decoupling of circulatory dynamics and neural regulatory mechanisms, which in turn causes a significant decrease in the above ratio. Therefore, this ratio becomes an important dynamic indicator for identifying specific pattern processes of such signals. To achieve individualized judgment, the system establishes a baseline value for the sign consistency ratio in the stable breathing state at the initial stage of user monitoring. The baseline value is generated by taking a moving average of the ratio values ​​calculated within the stable period, and has a certain degree of adaptive ability and can To adapt to individual physiological differences and state changes, the baseline value is established in a way that avoids the adaptability problems caused by static settings and can be dynamically updated during long-term monitoring. In real-time detection, if the system finds that the symbol consistency ratio of the current time period is continuously lower than the baseline value minus the first preset threshold, and the decline continues for more than a set minimum duration, it is considered that a rhythm desynchronization state signal segment may have occurred. Among them, the preset threshold should be set to a proportional decrease in the baseline value rather than a fixed value, so as to enhance adaptability to different individuals. For example, the proportion can be selected between 20% and 30%. The setting logic is to sensitively identify specific pattern abnormalities of the signal while reducing the probability of misjudgment of normal physiological fluctuations.

[0034] To further improve the accuracy of the determination, after the above-mentioned preliminary determination event occurs, the system will also calculate the second-order difference values ​​of the amplitude difference sequence and the period difference sequence respectively, and determine whether their absolute values ​​exceed the morphological identification threshold. If the absolute value of the second-order difference value exceeds the threshold range, the event can be determined as an artifact event caused by cardiac factors and thus be excluded. The morphological identification threshold is set based on historical physiological data and is generally used to distinguish between relatively slow changes caused by respiratory effort and violent fluctuations caused by sudden events such as high-frequency signal disturbances from non-respiratory sources, thereby enhancing the system's ability to identify interference from non-respiratory factors. The present invention also introduces body motion detection logic, which continuously monitors the user's body motion state through an acceleration sensor. When the body motion amplitude is detected to exceed the preset body motion threshold, the system will suspend the determination process of the rhythm desynchronization state signal segment to prevent body motion artifacts from interfering with the detection results. This mechanism constructs a synergistic barrier between rhythm analysis at the signal level and interference monitoring at the physical behavior level, enhancing the practical adaptability of the present method in daily life environments.Furthermore, during the time period when no rhythm desynchronization state signal segment is detected, the system will continue to calculate the modulation energy of the amplitude difference sequence and the period difference sequence. The modulation energy is obtained by calculating the variance of the difference sequence in the time period, which is used to quantify the dynamic activity of the pulse wave signal. When the two modulation energy values ​​are simultaneously and continuously lower than the lower limit threshold of their respective energy baselines within the set time length, a specific signal segment representing the silence of the signal modulation energy is identified. During the time period when no rhythm desynchronization state signal segment is determined to have occurred, the system continues to calculate the modulation energy of the amplitude difference sequence and the period difference sequence. In order to ensure that the identification of the specific signal segment representing the silence of the signal modulation energy has a clear engineering boundary, the duration can be set to a time length of not less than a complete respiratory rhythm cycle, usually corresponding to a sliding analysis window of thirty to sixty seconds. The selection of this time length is based on the fact that the specific signal segment representing the silence of the signal modulation energy has obvious continuous silence characteristics in physiological manifestations. Considering the mechanism, during this period, the loss of respiratory central drive causes the airflow, respiratory movement and blood flow modulation to tend to be silent, which in turn causes a double significant attenuation of the modulation energy in the pulse wave morphology; if within the above-mentioned time window, the modulation energy values ​​of the amplitude difference sequence and the period difference sequence are continuously kept below the lower limit threshold of the energy baseline, and this state continues without interruption until the end of the window, the system determines that the signal segment is the occurrence interval of a specific signal segment representing the silence of the signal modulation energy; this judgment logic is based on the characteristic that central respiratory apnea causes complete interruption of respiratory drive, which in turn manifests as a decrease in the silence of the signal, so low modulation energy becomes an important indicator factor for this type of event; the lower limit threshold of the modulation energy is set according to the modulation energy level corresponding to the user in a normal and stable sleep state, usually with 10% to 30% of this level as the lower limit of the proportion. At the same time, the energy baseline value can also be adaptively generated by moving average in the early stage of monitoring to ensure that it has the judgment ability that matches the individual characteristics of the user.

[0035] During the rhythm analysis of differential sequences, the system calculates the sign consistency ratio based on the correspondence between the amplitude difference sequence and the period difference sequence within a time window. The system then performs a signed product of the elements at the same moment and constructs a ratio indicator based on the statistical results of positive signs, thereby quantifying the synergy between the two in the direction of rhythmic fluctuations. During this process, each sequence element is labeled with a time series number to ensure the uniqueness of their one-to-one correspondence, thereby avoiding the risk of misjudgment due to sequence misalignment. The logic for setting this ratio is not based solely on mathematical operations but also on the physiological coupling characteristics of the pulse wave signal under respiratory drive. Under normal conditions, the pulse wave amplitude and period modulation tend to rise or fall synchronously. However, when the signal is in a specific mode or when the respiratory rhythm is disordered, inconsistent decoupling behavior often occurs, resulting in a significant decrease in the sign consistency ratio. Furthermore, the system introduces a dynamic trade-off mechanism for setting the first preset threshold. The fundamental technical consideration for this threshold is to achieve the optimal engineering balance between the ability to sensitively identify rhythm decoupling states and the overall anti-interference performance of the system. If the threshold is set too high, the system may miss invalid respiratory effort signals in specific patterns of mild signals. Conversely, if the threshold is too low, it may be overly sensitive to short-term noise or body motion disturbances, resulting in an increased false alarm rate. Therefore, in actual deployment, the threshold is selected based on the baseline symbol consistency ratio established under stable respiratory conditions at the beginning of monitoring. The relative judgment threshold is set by the proportional reduction, typically ranging from 20% to 30% of the baseline value, to ensure that the algorithm has sufficient recognition sensitivity while also taking into account judgment stability. To further improve the system's stability under low signal-to-noise ratio conditions, the signal acquisition module introduces a multi-scale smoothing mechanism and a pseudo-peak correction strategy. Specifically, during the pulse wave peak extraction process, local extreme value judgment is combined with amplitude threshold control. The threshold value is set at 50% of the average peak-to-valley difference within the resting segment to filter out high-frequency pseudo-peaks. At the same time, by jointly judging the time intervals and amplitude gradients of adjacent peaks, a peak legitimacy screening mechanism is established to ensure that the pulse wave peak sequence has good time alignment and representativeness at the input end. The introduction of the above strategy constructs a robust acquisition link based on signal stability and constrained by feature continuity, providing reliable support for subsequent differential calculation and rhythm analysis; for key parameters such as the symbol consistency ratio, the second-order difference value of the amplitude difference sequence and the period difference sequence, and the energy baseline lower limit threshold, the system does not use absolute constant settings. Instead, it combines the steady-state data obtained in the individual's initial monitoring phase and adaptively generates them through a sliding average and proportion setting mechanism. This method of establishing a parameter benchmark based on individual dynamic characteristics avoids the adaptability defects caused by static settings, while allowing the system to be iteratively optimized according to actual monitoring data during long-term operation, thereby constructing a highly robust judgment framework with endogenous feedback regulation capabilities.

[0036] Example 2: This example discloses a specific implementation process of a portable blood oxygen signal OSA intelligent detection method. The process is based on the photoelectric volumetric pulse wave signal collected by a single blood oxygen sensor, and is intended to achieve reliable detection of obstructive respiratory effort and specific signal segments that characterize signal modulation energy silence without relying on additional sensors. The core of the method is: by analyzing the rhythm synchronization between the pulse wave amplitude and the periodic differential signal, the invalid respiratory effort process is identified in time; and combined with morphological feature analysis and body motion monitoring information, artifact recognition and noise suppression are achieved, thereby achieving accurate judgment with low computing resources on the edge computing platform; the entire method includes the following processing stages: First, the system uses a photoelectric volumetric sensor to collect pulse wave signals. The typical wearing position of the sensor is the fingertip or earlobe. The collected original signal is first processed by Multi-scale smoothing is performed to improve signal stability; then, the pulse wave peak is extracted through a strategy based on neighborhood extreme value comparison and amplitude threshold constraint to form a peak sequence. This sequence serves as the basis for subsequent differential analysis, and its stability directly affects the accuracy of the overall judgment; based on the peak sequence, the system generates two types of differential sequences in parallel: one is the amplitude differential sequence, which is used to reflect the amplitude changes between adjacent pulse wave peaks; the other is the period differential sequence, which is used to characterize the fluctuations in the time intervals between consecutive heartbeats. These two types of differential sequences reveal the local dynamic characteristics of the pulse wave from the amplitude domain and time domain of the signal respectively. Within the set sliding time window, the system calculates the number of times the product of the corresponding elements of the above two types of differential sequences is positive, and compares it with the total number of heartbeats in the time period to obtain the sign consistency ratio as a quantitative indicator of rhythm synchronization.

[0037] In order to establish an individualized judgment benchmark, the system automatically selects a stable breathing period data when the user is in the initial resting state wearing the device, and obtains the symbol consistency ratio within this period by moving average, thereby constructing the user's baseline value. In the subsequent real-time monitoring process, if the current symbol consistency ratio is continuously lower than the relative threshold of the baseline value minus the preset ratio (such as 20%), and continues to exceed the set minimum time length, the system determines that it is a rhythm desynchronization state signal segment. This judgment mechanism uses the relative ratio form to set the threshold, which is conducive to adapting to the physiological differences between individuals and improving the versatility and practicality of the system. In order to further eliminate waveform abnormal interference caused by non-respiratory factors, the system performs second-order difference calculations on the amplitude difference sequence and the period difference sequence after identifying the suspected event. If any second-order difference sequence If the absolute value exceeds the morphological identification threshold, it is considered that the event may be caused by cardiac factors, and it is determined to be an artifact and eliminated. The above-mentioned morphological identification threshold is set based on the physiological fluctuation law disclosed in this field, and is often used to distinguish between slowly changing signals caused by respiratory efforts and sudden changes caused by abnormal heart rhythms; taking into account the signal disturbances that may be caused by changes in user body position during sleep at night, the system further integrates an acceleration sensor to monitor body movement status; if the body movement amplitude is detected to exceed the preset threshold, the threshold is set according to the range of body movement amplitude changes in typical wearing parts, the system will temporarily suspend the judgment process of the rhythm desynchronization state signal segment, and resume the judgment operation after the body movement tends to stabilize. This processing mechanism realizes the coordinated control between signal layer rhythm analysis and physical layer interference suppression, and improves the anti-interference ability of the overall detection system in daily environments.

[0038] During the time period when no signal segment of rhythm desynchronization is determined to have occurred, the system will continue to evaluate the modulation activity of the differential signal. Specifically, the system calculates the variance of the amplitude difference sequence and the period difference sequence within the set time window to measure the degree of dynamic change of the pulse wave. When the modulation energy of both is lower than the preset lower limit of the ratio of their respective baseline values ​​(such as 10% to 30%) for a continuous period of time, the system will determine this state as a specific signal segment that represents the silence of the signal modulation energy. The above baseline value can be adaptively determined by the user's fluctuation level in the initial resting period to adapt to individual differences and improve judgment accuracy. Through the orderly coordination of the above stages, this method can achieve highly robust recognition of obstructive and central respiratory events based on a single sensor, avoiding dependence on traditional blood oxygen saturation lag indicators, and has strong real-time performance and recognition sensitivity.

[0039] Example 3: This example constructs a platform for portable blood oxygen signal acquisition and analysis, which includes the following main components: a PPG signal acquisition module, which uses a photoelectric capacitance sensor and is worn on the subject's fingertip through a finger clip design to continuously acquire the original photoelectric capacitance pulse wave signal. The sensor operating frequency can be adjusted according to actual needs to ensure the stability of signal acquisition; a data preprocessing and feature extraction module: This module processes the original photoelectric capacitance pulse wave signal to detect the pulse wave peak of the photoelectric capacitance pulse wave signal in real time, thereby obtaining the pulse wave peak sequence. This processing process can include multi-scale Smoothing and pseudo-peak correction strategies are designed to improve the robustness of pulse wave peak extraction under low signal-to-noise ratio conditions; differential sequence generation and rhythm synchronization analysis module, based on the pulse wave peak sequence, generates in parallel an amplitude differential sequence reflecting the change in pulse wave amplitude and a period differential sequence reflecting the change in pulse cycle interval. Within a preset time window, the module analyzes the rhythm synchronization of the amplitude differential sequence and the period differential sequence. The rhythm synchronization is obtained by calculating the sign consistency ratio of the elements of the amplitude differential sequence and the corresponding elements of the period differential sequence. The sign consistency ratio represents the degree of sign consistency between the two sequences. In this experiment, The time window can be set to include the duration of a typical respiratory modulation cycle to effectively capture the rhythm coupling relationship in the pulse wave. The symbol consistency ratio can be calculated by counting the number of times the product of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence is positive within a preset time window, and the ratio of the number to the total number of heartbeats within the time window is used as the symbol consistency ratio. A reference value adaptive establishment and event judgment module compares the real-time symbol consistency ratio with a reference symbol consistency ratio. The reference symbol consistency ratio can be determined under a stable respiratory state at the beginning of the user monitoring. Specifically, the reference symbol consistency ratio can be adaptively established by calculating a moving average of the symbol consistency ratio under a stable respiratory state at the beginning of the user monitoring. When the real-time symbol consistency ratio is continuously lower than a first preset threshold of the reference symbol consistency ratio, the current signal segment can be determined to be a rhythm desynchronization state signal segment. The first preset threshold can be set as a proportional decrease in the reference value to enhance adaptability to different individuals. For example, the ratio can be selected between 20% and 30%. The setting logic is to reduce the probability of misjudging normal physiological fluctuations while sensitively identifying specific pattern abnormalities in the signal.Artifact identification and body motion monitoring module. This module continuously monitors the user's body motion through an acceleration sensor. When the body motion amplitude is detected to be greater than the preset body motion threshold, the judgment of the rhythm desynchronization state signal segment can be suspended to avoid the interference of body motion artifacts on the judgment of the rhythm desynchronization state signal segment. In addition, at the time point when the rhythm desynchronization state signal segment occurs, the second-order difference value of the amplitude difference sequence and the second-order difference value of the period difference sequence can be calculated. If the absolute value of the second-order difference value of the amplitude difference sequence or the absolute value of the second-order difference value of the period difference sequence is greater than the preset morphological identification threshold, the event can be marked as a high-frequency interference segment, and its weight can be reduced or excluded when calculating the risk index. The preset morphological identification threshold is set according to historical physiological data to distinguish between relatively gentle changes caused by respiratory efforts and instantaneous severe changes caused by high-frequency signal disturbances from non-respiratory sources. Changes: The central apnea determination module can continuously calculate the modulation energy of the amplitude difference sequence and the modulation energy of the period difference sequence during a time period in which a rhythm desynchronization signal segment is not determined to have occurred. The modulation energy is obtained by calculating the variance of the difference sequence within the time period. When the modulation energy of the amplitude difference sequence and the modulation energy of the period difference sequence are both continuously lower than their respective energy baseline lower limit thresholds for a preset duration, a specific signal segment representing a silence in the signal modulation energy can be identified. The energy baseline lower limit threshold can be set to a lower limit ratio of the corresponding modulation energy during the user's normal, stable sleep state, with the lower limit ratio value ranging from 0.1 to 0.3. The modulation energy baseline for normal, stable sleep can be adaptively established by taking a moving average of the variance of the amplitude difference sequence and the period difference sequence under the user's stable breathing state at the initial stage of monitoring.

[0040] This experiment aims to verify the ability of the portable blood oxygen signal OSA intelligent detection method to identify invalid respiratory effort events, identify artifacts, and distinguish different types of apnea. The experimental process is as follows: First, the photoelectric capacitance pulse wave signal is continuously collected through the photoelectric capacitance sensor. The collected original signal is processed by multi-scale smoothing to improve the signal stability. Then, the pulse wave peak of the photoelectric capacitance pulse wave signal is detected in real time through a strategy based on neighborhood extreme value comparison and amplitude threshold constraint to obtain the pulse wave peak sequence. The stability of the sequence directly affects the accuracy of subsequent judgments; based on the pulse wave peak sequence The amplitude difference sequence and the period difference sequence are generated in parallel. The amplitude difference sequence is used to reflect the amplitude change between adjacent pulse wave peaks, and the period difference sequence is used to characterize the time interval fluctuation between consecutive heartbeats. Within the preset time window, the number of times the product of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence is positive is calculated, and the ratio of the number to the total number of heartbeats in the time window is used as the symbol consistency ratio to obtain rhythm synchronization. This ratio quantifies the degree of consistency of the two difference sequences in the direction of symbol fluctuation. In the stable breathing state at the initial stage of user monitoring, the stability period is calculated by The symbol consistency ratio obtained is moved and averaged to generate a reference value. In real-time detection, if the symbol consistency ratio of the current time period continues to be lower than the first preset threshold of the reference symbol consistency ratio, the threshold can be set to a proportional decrease of 20% to 30% of the reference value, and the decrease state continues for more than a set minimum time, then the current signal segment is determined to be a rhythm desynchronization state signal segment. The preset threshold is intended to enhance adaptability to different individuals and reduce the probability of misjudgment of normal physiological fluctuations while sensitively identifying abnormalities in specific patterns of signals; after the preliminary determination of the occurrence of the rhythm desynchronization state signal segment Then, the second-order difference values ​​of the amplitude difference sequence and the period difference sequence are calculated respectively. If the absolute value of any second-order difference value is greater than the preset morphological identification threshold, the event is marked as a high-frequency interference segment and excluded. The morphological identification threshold can be set according to historical physiological data to distinguish between relatively slow changes caused by respiratory effort and violent fluctuations caused by sudden high-frequency signal disturbances from non-respiratory sources. At the same time, the user's body movement is continuously monitored by the acceleration sensor. When the body movement amplitude is detected to be greater than the preset body movement threshold, the determination of the rhythm desynchronization state signal segment is suspended to avoid interference of body movement artifacts on the detection results.During a time period where no rhythm desynchronization signal segments are detected, the modulation energy of the amplitude difference sequence and the period difference sequence is continuously calculated. The modulation energy is obtained by calculating the variance of the difference sequences within the time period. When both modulation energy values ​​are simultaneously and continuously below their respective energy baseline lower thresholds for a set period of time, a specific signal segment representing a silence in the signal modulation energy is determined to have occurred. The modulation energy lower threshold can be set based on the modulation energy level corresponding to the user's normal, stable sleep state, typically with a proportional lower limit of 10% to 30% of that level. Furthermore, the energy baseline value can also be adaptively generated using a moving average method during the initial monitoring phase.

[0041] This experiment focused on a detailed analysis of the performance of this method in identifying ineffective breathing efforts, distinguishing artifacts, and differentiating different types of apnea. The role of the sign consistency ratio in identifying ineffective breathing efforts was observed in this experiment. When the user had an ineffective breathing effort event, the sign consistency ratio showed a significant and sustained downward trend. When the user was in a normal breathing state, there was a relatively stable rhythm synchronization between the amplitude difference and the period difference, and the sign consistency ratio usually remained at a high level. When a rhythm desynchronization signal segment occurred, the airway obstruction caused abnormal fluctuations in the negative pressure in the lungs, resulting in rhythmic decoupling of the circulatory dynamics and the neural regulatory mechanism, and then This method can effectively distinguish cardiac artifacts from body motion artifacts, and effectively distinguish cardiac artifacts from body motion artifacts. During the experiment, this method demonstrated its ability to distinguish scenes where artifacts may be introduced. If the absolute value of the second-order difference value exceeds the morphological distinction threshold, The system can identify artifact events caused by cardiac factors. This morphological discrimination threshold is designed to distinguish between relatively slow-changing phenomena caused by respiratory efforts and violent fluctuations caused by sudden high-frequency signal disturbances from non-respiratory sources. In addition, when the body movement amplitude is detected to exceed the preset body movement threshold, the system can suspend the judgment process of the rhythm desynchronization state signal segment, which effectively avoids the interference of body movement artifacts on the detection results, and builds a synergistic barrier between rhythm analysis at the signal level and interference monitoring at the physical behavior level, which significantly enhances the actual adaptability of this method in daily life; To distinguish between obstructive and central apnea, this method can distinguish by monitoring the abnormal attenuation of differential sequence modulation energy. The system can identify specific signal segments that represent the silence of signal modulation energy. When the modulation energies of the amplitude difference sequence and the period difference sequence are both continuously lower than the respective energy baseline lower thresholds for a preset duration within a time period in which no signal segment of rhythm desynchronization has occurred, a specific signal segment that represents the silence of signal modulation energy can be identified. This judgment logic is based on the characteristic that central apnea leads to a complete interruption of respiratory drive, which in turn manifests as a decrease in the silence of the signal. Therefore, low modulation energy becomes an important indicator factor for this type of event. This capability enables a single blood oxygen sensor to distinguish between obstructive and central apnea for the first time, providing a key classification basis for precise treatment decisions.

[0042] Example 4: This example combines Figures 1 to 3 , a portable blood oxygen signal OSA intelligent detection method is described. Figure 1As shown in the figure, the accelerometer outputs three-axis acceleration data in real time to the main processor for extraction and analysis of body motion amplitude values. When the body motion amplitude meets the amplitude <0.5G condition, the main processor enters the alt state and triggers the activation of respiratory analysis. During this period, the system marks the current data segment as a valid signal mark and hands it over to the event analysis module for further processing. If the amplitude is detected to be ≥0.5G, it enters the high body motion state, and the system will pause the event judgment operation and synchronously cache the original data for subsequent resumption. After the body motion event ends, the system enters the recovery monitoring phase. If the user is detected to be still for 2 seconds and returns to a stable signal state, the main processor will perform restart analysis and resume marking, resuming the analysis and judgment of respiratory events. The figure also clearly sets the body motion threshold: 0.5G and pause delay: 200ms, which are used to trigger and control the boundary conditions of each processing logic. This process effectively avoids the influence of artifact interference on the judgment of rhythm desynchronization state signal segments by intelligently pausing and resuming the respiratory analysis process under body motion interference conditions, ensuring the accuracy and robustness of the detection system.

[0043] like Figure 2 As shown, Figure 2 (a) A line graph shows the change in rhythm synchrony ratio over monitoring time (hh:mm), with states A, B, and C used to mark different event types. This graph shows that between 00:00 and 03:00, the rhythm synchrony ratio gradually decreases from an initial level close to 1.0, then decreases significantly between 01:30 and 02:30, before recovering to approximately 0.9 at 03:00. This reflects the fluctuating nature of the synchrony between the amplitude difference sequence and the period difference sequence in the pulse wave, highlighting the trend of rhythm decoupling in the early stages of respiratory events in states B and C. The dashed baseline in the figure is used to determine the trigger threshold for decreased synchrony, and the gray background emphasizes the range of variation of this indicator. Figure 2 The bar graph in (b) shows the distribution of amplitude modulation activity values ​​in each sleep stage at the same monitoring time, including the awake period, N1 period, N2 period, N3 period and REM period. From 00:00 to 03:00, the activity level showed an overall downward trend, especially at 02:30 (REM period), where it dropped to the lowest level, confirming the correlation between respiratory drive quiescence and state C pause.

[0044] like Figure 3As shown, first, by continuously collecting and extracting the peak value of the photoelectric volume pulse wave signal, the following are generated in parallel: amplitude difference sequence and period difference sequence, to construct the basic feature data for subsequent analysis. The system establishes a baseline symbol consistency ratio under a stable respiratory state, and forms an individualized baseline by moving the average baseline. Then, the symbol consistency ratio (amplitude and period difference symbol matching) of the differential sequence is calculated and compared with the baseline value. If the real-time symbol consistency ratio is lower than the baseline threshold, the rhythm desynchronization state signal segment is determined, and the abnormal analysis stage is entered. In order to improve the accuracy of the judgment and reduce misjudgment, the system introduces cardiac artifact exclusion, which specifically includes two-order differential morphological analysis and body motion monitoring pause logic, so as to effectively identify high-frequency signal disturbances from non-respiratory sources and interference caused by body motion artifacts. Under the premise of not triggering an obstructive event, the system also monitors the modulation activity. If the modulation energy is continuously lower than the baseline, a specific signal segment representing the silence of the signal modulation energy is identified, thereby constructing a highly robust respiratory event recognition process based on the triple judgment mechanism of rhythm synchronization, morphological change and energy characteristics, ensuring the accuracy and adaptability of the technical solution in the edge device environment. As shown Figure 4 As shown in Figure 1, the changing trend of the symbol consistency ratio during the 24-hour monitoring period is shown. The solid line represents the real-time symbol consistency ratio, the dotted line represents the baseline value, and the dotted line represents the judgment threshold (70% of the baseline value). When the real-time ratio is continuously lower than the threshold, the system determines it as a rhythm desynchronization state signal segment; as shown in Figure 1, the real-time symbol consistency ... Figure 5 As shown in the figure, the modulation energy changes of the amplitude difference sequence and the period difference sequence under different breathing states are shown. The solid line represents the modulation energy of the amplitude difference sequence, the long dashed line represents the modulation energy of the period difference sequence, and the dotted line represents the lower limit threshold of the energy baseline; Figure 6 The figure shows the relationship between body motion amplitude and respiratory analysis status. The upper part shows the change in body motion amplitude (solid line), and the lower part shows the system analysis status (square mark indicates normal analysis, and triangle mark indicates paused analysis).

[0045] Example 5: In the process of constructing the differential sequence of this embodiment, the system first adopts a pulse peak extraction method that combines neighborhood local extreme value judgment and amplitude threshold control based on the continuously collected photoelectric volume pulse wave signal, and cooperates with the pseudo-peak suppression strategy to generate a pulse wave peak sequence. In order to enhance the stability of heartbeat alignment, the system sets a sampling frequency standardization module in the peak extraction process and introduces a numbering marking mechanism to ensure sequence consistency in different heart rate intervals. On this basis, the system calculates the amplitude difference and period difference between adjacent peaks respectively to form an amplitude differential sequence and a period differential sequence for reflecting the change trend of the blood oxygen signal, and ensures that the two differential sequences have a one-to-one correspondence at each time node through number association. element relationship; in terms of the generation of rhythm synchronization indicators, the system introduces a sliding analysis window of fixed length to count the degree of rhythm coupling between differential sequences. The window length is usually set to thirty to forty seconds, covering several complete respiratory cycles. This setting comprehensively considers the timing characteristics of the respiratory rhythm and the stability of the algorithm response. It can not only cover representative signal change patterns, but also effectively suppress short-term abnormal interference. Within this time window, the system compares the signs of the products of corresponding elements in the amplitude difference sequence and the period difference sequence one by one, counts the number of times the product is positive, and forms a ratio with the total number of heart beats in the window. This ratio is the sign consistency ratio, which is used to quantify the consistency level of the respiratory regulation rhythm in the hemodynamic response.

[0046] In order to adapt to individual differences, the system selects the signal segment of the user in the resting state to construct the baseline value in the initial stage of wearing the device. In the specific process, the system uses multi-window smoothing calculation to obtain the symbol consistency ratio in the initial stage, and establishes statistical expectations by eliminating abnormal values ​​that deviate from the average level. It serves as the individual baseline symbol consistency ratio for subsequent event judgment. This method can achieve baseline self-learning without relying on additional physiological parameters, and is suitable for the diversity of basic states among users. In actual operation, the system continuously monitors the changes in the current symbol consistency ratio. If it is found that the ratio decreases by more than the set ratio of the baseline value in a continuous time period, and the duration reaches the system response threshold, the system determines that there is respiratory rhythm decoupling in this period, and combines other signal features to further determine whether it constitutes a rhythm desynchronization state signal segment. The ratio change threshold is usually set in the range of 25% to 30%. The system comprehensively considers the need to distinguish between the amplitude of rhythm fluctuations and abnormal states during normal sleep; in order to improve the accuracy of event recognition, the system also introduces a second-order difference calculation mechanism for differential sequences, performs second-order difference processing on the amplitude difference sequence and the period difference sequence respectively, and calculates the absolute value of the obtained value to evaluate the severity of the fluctuation. When a suspected respiratory event is detected, the system compares the second-order difference intensity of the current time period with the morphological threshold previously constructed in a resting state or a normal rhythm state. If the difference exceeds the threshold range, it indicates that there are abnormal and violent fluctuations in the signal, which may be caused by non-respiratory related reasons such as high-frequency signal disturbances from non-respiratory sources. The system then determines the event as a non-respiratory artifact event and excludes it from the final event record. The setting principle of the morphological threshold is based on the changing characteristics of typical heart waveform mutations, taking into account the influence of common noise types on blood oxygen signals, and maintaining its dynamic adaptability through a smoothing update mechanism.

[0047] During the detection of central apnea, the system evaluates the fluctuation intensity of the amplitude difference sequence and the period difference sequence, and uses a sliding window to count the respective signal variances. The calculated variance value is used to reflect the degree of blood oxygen signal modulation activity during the time period. If the variance value is lower than the preset energy baseline lower limit in several consecutive time windows, it is regarded as a modulation energy silent state. The system recognizes that there may be a specific signal segment that represents the signal modulation energy silence; the above-mentioned energy baseline lower limit is usually based on the resting state samples obtained during the initial operation of the device. The threshold is set by calculating the average variance level and selecting the 10% to 30% range. The selection of the setting range is combined with the signal noise level and the system false alarm tolerance to make it adaptable between different users; in terms of body motion interference judgment, the system collects the three-axis acceleration signal of the wearing part in real time, calculates the current body motion amplitude by vector synthesis, and combines it with the judgment time A dynamic interference recognition model is constructed based on the time length. When the body motion amplitude value continuously exceeds 0.5 gravity acceleration units and lasts for more than 200 milliseconds, the system enters a high-interference state and temporarily suspends the respiratory event judgment process. All signal data will enter the cache area. After that, if the system detects that the acceleration signal continuously returns to a static state and remains for more than two seconds, the event analysis process is re-enabled and the cache area data is included in the subsequent processing. The above judgment conditions adopt a dual judgment mechanism of amplitude threshold and time threshold to ensure good anti-interference ability for non-respiratory related factors such as night-time body position adjustment and slight shaking, further ensuring the continuity and accuracy of the analysis; through the timing coordination, data intercommunication and logical linkage between the above modules, the system realizes the coordinated detection of obstructive and central sleep respiratory events under the condition of a single blood oxygen signal input, ensuring its robustness and feasibility in different usage environments and user states.

[0048] Example 6: This example aims to elaborate on the specific calculation process of the core indicator of the present invention, namely, the symbol consistency ratio. The entire calculation process can be decomposed into the following steps: Step 1: Obtain a pulse wave peak sequence. The pulse wave signal is continuously collected by a photoplethysmography (PPG) sensor, and its peak value is detected in real time to obtain a pulse wave peak sequence. The sequence can be formally expressed as : ,in, Representative The information of the pulse wave peak point. : No. The peak amplitude of the pulse wave. : No. The time when the peak of the pulse wave occurs.

[0049] Step 2: Generate amplitude difference sequence and period difference sequence based on pulse wave peak sequence , the system generates the differential sequences of the two cores in parallel. Amplitude differential sequence ( ): This sequence reflects the change in peak amplitude between consecutive pulse waves. Elements Defined as: , periodic difference sequence ( ): This sequence reflects the changes in the intervals between consecutive heartbeat cycles. First, the pulse cycle sequence needs to be calculated based on the peak time series. , the first Elements Defined as the time interval between two adjacent peaks: , then, the periodic difference sequence No. Elements Defined as the difference between two adjacent pulse cycles: , by the above definition, the elements of the amplitude difference sequence Elements of the periodic difference sequence By common heartbeat index number A one-to-one correspondence is formed, laying the foundation for subsequent synchronization analysis.

[0050] Step 3: Calculate the symbol consistency ratio ( ), within a preset time window (e.g., 30-40 seconds, the symbol consistency ratio Calculation, statistics within the window, the elements of the amplitude difference sequence Corresponding elements of periodic difference sequence The number of times the product sign is positive, and the ratio of this number to the total number of heartbeats in the window is taken as the final result. The calculation formula can be expressed as: ,in, is the total number of heartbeats within the preset time window; : is the index of the difference sequence, from 1 to ; : is a counting function. When the condition in the brackets ( ) is true, its value is 1; when the condition is false, its value is 0. The symbol consistency ratio proposed by the present invention The conception of this method stems from a profound insight into the specific physiological processes of obstructive sleep apnea (OSA) signals, and aims to address the fundamental limitation of existing technologies that can only passively rely on delayed obstruction results such as decreased blood oxygen saturation for judgment. The core principle is that during normal and stable breathing, the human body's respiratory movement produces synchronous modulation on the amplitude and period of the photoplethysmogram (PPG) signal through changes in intrathoracic pressure and autonomic nerve regulation, which manifests as a physiological coupling of the changing trends of the two. When OSA patients have airway obstruction and make ineffective breathing efforts, the severe and irregular fluctuations in intrathoracic negative pressure will destroy this regulatory pathway, resulting in a significant decoupling of the amplitude and period changes of the pulse wave. This formula is designed to quantitatively capture this transition process from coupling to decoupling: by calculating the amplitude difference sequence ( ) and the periodic difference sequence ( ) to determine whether the change directions of the two are consistent, thereby identifying the heartbeat in the coupled state; the final The ratio, that is, the proportion of coupled heartbeats to the total heartbeats, constructs a new OSA detection dimension that is independent of blood oxygen saturation. This formula can directly perceive the specific pattern process of the core signal in the early stage of obstruction, thereby achieving effective screening for mild to moderate patients.

[0051] The physical meaning of this formula is to quantify the degree of synchronization between the amplitude change trend and the period change trend. The higher the value, the better the synchronization and the smoother the breathing; otherwise, it means that rhythm decoupling has occurred, which may indicate the occurrence of ineffective breathing efforts.

[0052] Step 4: To make the calculation process of the present invention clearer, a simplified numerical example is provided. Assume that within a time window, the system collects 6 consecutive pulse wave peaks and the total number of heartbeats is See Table 1, raw data (corresponding to step 1).

[0053] Table 1: Example table of raw data of pulse wave peak sequence.

[0054] Intermediate sequence calculation (corresponding to step 2), calculate the pulse cycle : s; s; s; s; s; Calculate the difference sequence and : Table 2 below shows the index arrive The calculation result (because Need to use ).

[0055] Table 2: Example table of differential sequence calculation results.

[0056] Table 3: Example table of symbol consistency determination process.

[0057] Final ratio calculation: According to the formula, the numerator is the sum of the times the product is positive (3 times), and the denominator is the total number of heartbeats in the window (6 times), The calculated real-time symbol consistency ratio (0.5 in this example) is fed into the event determination module and compared with a baseline symbol consistency ratio established through a moving average. If this value remains below the baseline threshold, the system determines that a rhythm desynchronization signal segment has occurred.

[0058] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A portable blood oxygen signal OSA intelligent detection method, characterized in that: The method comprises the following steps: Step a, continuously collecting photoplethysmography signals and detecting the pulse wave peak value of the photoplethysmography signals in real time to obtain a pulse wave peak sequence; Step b, based on the pulse wave peak sequence, generating in parallel an amplitude difference sequence reflecting the change of the pulse wave amplitude and a period difference sequence reflecting the change of the pulse cycle interval; Step c: analyzing the rhythmic synchronization of the amplitude difference sequence and the period difference sequence within a preset time window. The rhythmic synchronization is obtained by calculating the sign consistency ratio of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence. The sign consistency ratio represents the degree of sign consistency of the two sequences. Step d, comparing the symbol consistency ratio obtained in real time with a reference symbol consistency ratio, where the reference symbol consistency ratio is determined under a stable breathing state at the initial stage of user monitoring; Step e: when the real-time symbol consistency ratio is continuously lower than the first preset threshold value of the reference symbol consistency ratio, the current signal segment is determined to be a rhythm desynchronization state signal segment.

2. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: In step c, the symbol consistency ratio is calculated as follows: within a preset time window, count the number of times the product of the elements of the amplitude difference sequence and the corresponding elements of the period difference sequence is positive, and use the ratio of the number to the total number of heartbeats in the time window as the symbol consistency ratio.

3. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: The reference symbol consistency ratio is adaptively established by calculating a moving average of the symbol consistency ratio in a stable breathing state at the initial stage of user monitoring.

4. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: The following steps are also included: When the cumulative duration or occurrence frequency of the determined rhythm desynchronization state signal segment reaches or exceeds a preset reference threshold, a high-risk level warning is output.

5. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: After determining in step e that a rhythm desynchronization state signal segment has occurred, the following steps are also included: calculating the second-order differential value of the amplitude difference sequence and the second-order differential value of the period difference sequence at the time point when the rhythm desynchronization state signal segment occurs; if any of the absolute values ​​of the second-order differential values ​​of the amplitude difference sequence or the absolute values ​​of the second-order differential values ​​of the period difference sequence is greater than a preset morphological identification threshold, the event is marked as a high-frequency interference segment, and its weight is reduced or excluded when calculating the risk index.

6. The portable blood oxygen signal OSA intelligent detection method according to claim 5, characterized in that: The preset morphological discrimination threshold is set based on historical physiological data to distinguish relatively gentle changes caused by respiratory effort from instantaneous and drastic changes caused by high-frequency signal disturbances from non-respiratory sources.

7. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: The method also includes the following steps: continuously monitoring the user's body movement through an acceleration sensor; and when the body movement amplitude is detected to be greater than a preset body movement threshold, suspending the judgment of step e to avoid interference of body movement artifacts in the judgment of the rhythm desynchronization state signal segment.

8. The portable blood oxygen signal OSA intelligent detection method according to claim 1, characterized in that: The following steps are also included: During the time period when the rhythm desynchronization state signal segment is not determined to have occurred in step e, the modulation energy of the amplitude difference sequence and the modulation energy of the period difference sequence are continuously calculated, where the modulation energy is obtained by calculating the variance of the difference sequence within the time period; When the modulation energy of the amplitude difference sequence and the modulation energy of the periodic difference sequence At the same time, when the energy is continuously lower than the respective energy baseline lower limit threshold within a preset duration, a specific signal segment representing the silence of the signal modulation energy is identified.

9. The portable blood oxygen signal OSA intelligent detection method according to claim 8, characterized in that: The energy baseline lower limit threshold is set to the lower limit ratio value of the corresponding modulation energy in the user's normal and stable sleep state, and the range of the lower limit ratio value is 0.1 to 0.

3.

10. The portable blood oxygen signal OSA intelligent detection method according to claim 8, characterized in that: The modulation energy baseline in a normal stable sleep state is adaptively established by taking a moving average of the variances of the amplitude difference sequence and the period difference sequence under a stable breathing state at the initial stage of user monitoring.

Images