Sleep apnea monitoring method, device and equipment and storage medium

Abstract

The invention provides a sleep apnea monitoring method, device and equipment and a storage medium. According to the sleep apnea monitoring method and device, shallow and deep physiological data are captured through multiple wavelength signals corresponding to multiple light sources, and blood oxygen signals are obtained through the obtained physiological data; then, analyzing local features and different weights of various physiological data and blood oxygen signals to obtain global features representing a change time sequence of a sleep breathing event of the user, and judging whether the user has a hypopnea event or not by analyzing continuous fragment features of a first preset duration in the global features; according to the scheme, whether the hypoventilation event is accurate or not is verified again in a mode of secondary confirmation of the abnormal signals according to the condition that whether the hypoventilation event occurs or not and whether continuous abnormal signals still occur or not after the hypoventilation event occurs, and the scheme can be completed through the wearable equipment, so that the equipment has the advantages of miniaturization, low power consumption and low cost; and the technical scheme is obtained through comprehensive analysis of various data, so that the accuracy of the result can be improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method, apparatus, device, and storage medium for monitoring sleep apnea. Background Technology

[0002] Sleep apnea syndrome (SAS) is a common sleep disorder characterized by recurrent episodes of apnea or hypopnea during sleep. This condition not only affects sleep quality but is also a significant risk factor for various chronic diseases such as hypertension, heart disease, and stroke. Therefore, early screening and diagnosis of sleep apnea are of great importance.

[0003] Currently, clinical diagnosis of sleep apnea largely relies on polysomnography (PSG). This technique requires a hospital or specialized sleep center and involves simultaneously recording dozens of physiological parameters, including electroencephalography (EEG), electrooculography (EOG), electrocardiography (ECG), electromyography (EMG), nasal and oral airflow, blood oxygen saturation, and respiratory effort, using numerous electrodes and sensors connected to the patient's head, face, chest, abdomen, and limbs. Although PSG is highly accurate, its process is complex and costly. Furthermore, it requires patients to fall asleep in an unfamiliar environment, resulting in poor comfort and low accessibility, making it unsuitable for routine initial screening and long-term monitoring.

[0004] Therefore, there is an urgent need in this field for a solution that can ensure high detection accuracy while also taking into account device miniaturization, low power consumption, and low cost. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, device, and storage medium for monitoring sleep apnea, to provide a solution that ensures high detection accuracy while also achieving device miniaturization, low power consumption, and low cost.

[0006] In a first aspect, embodiments of this application provide a method for monitoring sleep apnea, the method comprising: Acquire PPG signals of multiple color light sources collected by wearable devices at night; The user's blood oxygen signal is obtained from the red light PPG signal and the infrared light PPG signal in the PPG signals of multiple color light sources; Based on the local features of PPG signals and blood oxygen signals from multiple color light sources, and the feature matrix constructed with weights configured for different signals, global features for characterizing the temporal sequence of changes in sleep breathing events in users are obtained. Classify the continuous segment features of the first preset duration in the global features, and determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypoventilation event; When a hypoventilation event occurs, it is determined whether the blood oxygen signal exhibits an abnormal signal exceeding a second preset duration within the first preset duration; If an abnormal signal occurs for more than the second preset duration, the hypoventilation event is confirmed to have actually occurred, indicating that the user experienced sleep apnea at night.

[0007] In one feasible implementation, obtaining the user's blood oxygen signal based on the red PPG signal and infrared PPG signal from multiple color light sources includes: Based on the timestamps of the red PPG signal and the infrared PPG signal, obtain the red PPG signal segment and the infrared PPG signal segment of the third preset duration. For red PPG signal segments and infrared PPG signal segments, the peak value and valley value of the red PPG signal segments and infrared PPG signal segments within each sliding window are determined using preset sliding windows and steps, respectively. Based on the peak and trough values ​​of the red PPG signal segment and the infrared PPG signal segment within each sliding window, calculate the blood oxygen value corresponding to each sliding window. The blood oxygen values ​​are interpolated according to a preset sampling frequency to obtain the blood oxygen signal.

[0008] In one feasible implementation, a feature matrix is ​​constructed based on the local characteristics of PPG signals and blood oxygen signals from multiple color light sources, and the weights configured for different signals, including: The PPG signals of various color light sources and the blood oxygen signal are subjected to maximum and minimum normalization processing respectively to obtain the first value of the PPG signal of each color light source and the second value of the blood oxygen signal. Based on the first value and the second value, determine the first local feature of each first value and the second local feature of each second value; Using a global attention mechanism, calculate the first weight of each first local feature and the second weight of each second local feature; The feature matrix is ​​obtained by concatenating the product of the first local feature and its corresponding first weight, and the product of the second local feature and its second weight.

[0009] In one feasible implementation, obtaining the global features used to characterize the temporal sequence of changes in a user's sleep breathing events includes: The global features are obtained by extracting features from the feature matrix using a temporal feature model. In a feasible implementation, classifying continuous segment features of a first preset duration in the global features to determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypopnea event includes: The continuous segment features are classified using a classification model to determine the first probability of a normal breathing event and the second probability of a hypoventilation event. When the second probability is greater than the first probability, it is determined that the user corresponding to the continuous segment feature has a hypoventilation event during sleep.

[0010] In one feasible implementation, the method further includes: The number of sleep apnea episodes per unit time period is determined, and this number is used as the sleep apnea-hypopnea index.

[0011] In one feasible implementation, the wearable device includes: a microcontroller, a time-division driving circuit, multiple color light sources, and a photoelectric sensor; acquiring the PPG signals of the multiple color light sources collected by the wearable device at night includes: The microcontroller controls the time-division drive circuit to use time-division multiplexing to sequentially light up the corresponding color light source according to the preset cycle sequence, and the photoelectric sensor obtains the current signal of the corresponding color light source. According to the preset sampling frequency, the current signals of each color light source are digitally processed to obtain the original PPG signals of multiple color light sources. The original PPG signals of each color light source are low-pass filtered to obtain PPG signals of multiple color light sources.

[0012] Secondly, embodiments of this application provide a sleep apnea monitoring device, the device comprising: The acquisition unit is used to acquire PPG signals of various color light sources collected by the wearable device at night; The first determining unit is used to obtain the user's blood oxygen signal based on the red light PPG signal and the infrared light PPG signal from the PPG signals of multiple color light sources; The second determining unit is used to obtain global features that characterize the temporal sequence of changes in the user's sleep breathing events based on the local features of PPG signals and blood oxygen signals from multiple color light sources, as well as the feature matrix constructed by weights configured for different signals. The third determining unit is used to classify the continuous segment features of the first preset duration in the global features and determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypoventilation event. The fourth determining unit is used to determine whether the blood oxygen signal has an abnormal signal exceeding a second preset duration within the first preset duration when a hypoventilation event occurs. The fifth determining unit is used to determine that the hypoventilation event has actually occurred when an abnormal signal exceeding the second preset duration is detected, thus confirming that the user has experienced sleep apnea at night.

[0013] In one feasible implementation, when the first determining unit obtains the user's blood oxygen signal based on the red PPG signal and the infrared PPG signal from multiple color light sources, it includes: Based on the timestamps of the red PPG signal and the infrared PPG signal, obtain the red PPG signal segment and the infrared PPG signal segment of the third preset duration. For red PPG signal segments and infrared PPG signal segments, the peak value and valley value of the red PPG signal segments and infrared PPG signal segments within each sliding window are determined using preset sliding windows and steps, respectively. Based on the peak and trough values ​​of the red PPG signal segment and the infrared PPG signal segment within each sliding window, calculate the blood oxygen value corresponding to each sliding window. The blood oxygen values ​​are interpolated according to a preset sampling frequency to obtain the blood oxygen signal.

[0014] In one feasible implementation, the second determining unit, when constructing a feature matrix based on the local features of PPG signals and blood oxygen signals from multiple color light sources, and the weights configured for different signals, includes: The PPG signals of various color light sources and the blood oxygen signal are subjected to maximum and minimum normalization processing respectively to obtain the first value of the PPG signal of each color light source and the second value of the blood oxygen signal. Based on the first value and the second value, determine the first local feature of each first value and the second local feature of each second value; Using a global attention mechanism, calculate the first weight of each first local feature and the second weight of each second local feature; The feature matrix is ​​obtained by concatenating the product of the first local feature and its corresponding first weight, and the product of the second local feature and its second weight.

[0015] In one feasible implementation, when the second determining unit obtains global features characterizing the temporal sequence of changes in a user's sleep breathing events, it includes: The global features are obtained by extracting features from the feature matrix using a time-series feature model.

[0016] In a feasible implementation, the third determining unit is used to classify continuous segment features of a first preset duration in the global features, and to determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypopnea event, including: The continuous segment features are classified using a classification model to determine the first probability of a normal breathing event and the second probability of a hypoventilation event. When the second probability is greater than the first probability, it is determined that the user corresponding to the continuous segment feature has a hypoventilation event during sleep.

[0017] In one feasible implementation, the device further includes: The sixth determining unit is used to determine the number of sleep apnea events occurring per unit time, and to use the number of events as the sleep apnea-hypopnea index.

[0018] In one feasible implementation, the wearable device includes: a microcontroller, a time-division driving circuit, multiple color light sources, and a photoelectric sensor; The acquisition unit is used to acquire PPG signals of multiple color light sources collected by the wearable device at night, including: The microcontroller controls the time-division drive circuit to use time-division multiplexing to sequentially light up the corresponding color light source according to the preset cycle sequence, and the photoelectric sensor obtains the current signal of the corresponding color light source. According to the preset sampling frequency, the current signals of each color light source are digitally processed to obtain the original PPG signals of multiple color light sources. The original PPG signals of each color light source are low-pass filtered to obtain PPG signals of multiple color light sources.

[0019] Thirdly, embodiments of this application provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as described in any one of the first aspects.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the method as described in any one of the first aspects.

[0021] The technical solution provided in this application includes, but is not limited to, the following beneficial effects: In this application, multiple wavelength signals corresponding to various light sources are used to capture shallow and deep physiological data, and blood oxygenation signals are obtained from the acquired physiological data. Then, by analyzing the local features and different weights of various physiological data and blood oxygenation signals, global features characterizing the temporal sequence of changes in the user's sleep breathing events are obtained. By analyzing the features of continuous segments of the first preset duration in the global features, it is determined whether the user has experienced a hypopnea event, and whether there are any continuous abnormal signals after the hypopnea event. The accuracy of the hypopnea event is verified again by a secondary confirmation of abnormal signals, that is, to verify whether the user actually experienced sleep apnea at night. Since the above solution can be completed by wearable devices, the devices have the advantages of miniaturization, low power consumption and low cost. Furthermore, since the technical solution is obtained through comprehensive analysis of multiple data, it is beneficial to improve the accuracy of the results.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic flowchart illustrating a sleep apnea monitoring method provided in this application embodiment; Figure 2 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment; Figure 3 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment; Figure 4 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment; Figure 5 This is a schematic diagram of the structure of a sleep apnea monitoring device provided in an embodiment of this application; Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0026] With the development of wearable device technology, sleep monitoring technology based on photoplethysmography (PPG) signals has gradually emerged, making it possible to screen for sleep apnea at home in a convenient manner. Existing PPG-based technical solutions mainly include the following categories: (1) Detection method based on single-wavelength PPG signal: This method usually uses a single-wavelength (such as green or red light) PPG sensor to collect pulse wave signals. By analyzing the waveform amplitude changes and heart rate variability of the PPG signal, the occurrence of apnea event can be indirectly inferred.

[0027] (2) Detection method based on multi-physiological parameter fusion: To improve accuracy, existing technologies attempt to combine PPG signals with data from multiple sensors such as accelerometers and independent blood oxygen saturation sensors. Respiratory events are comprehensively judged through multi-sensor data fusion algorithms.

[0028] (3) Detection methods based on traditional machine learning: These methods first manually extract various features such as time domain and frequency domain from PPG signals, and then use traditional machine learning classifiers such as support vector machine (SVM) and random forest to perform pattern recognition and determine the sleep breathing state.

[0029] Despite the progress made by the above three technologies, there are still obvious limitations in practical applications: Insufficient signal quality and information richness: Single-wavelength PPG signals are easily affected by factors such as motion artifacts, ambient light interference, and changes in wearing position, resulting in unstable signal quality. At the same time, the physiological information carried by a single wavelength (such as green light, which mainly reflects superficial blood flow, while red / infrared light can penetrate deeper tissues) is limited, and it is impossible to fully utilize the specific responses of different wavelengths of light to different physiological parameters such as blood volume and blood oxygenation, leading to insufficient basis for judging respiratory events.

[0030] Limitations of feature extraction capabilities: Traditional machine learning methods rely heavily on expert experience for manual feature design and selection, making it difficult to comprehensively and automatically capture deep, temporal patterns related to complex respiratory events in PPG signals, and the model performance has a ceiling.

[0031] The system is complex and costly: the solution relies on the fusion of multiple sensors (such as PPG + accelerometer + independent blood oxygen probe), which increases the complexity of hardware design, device power consumption and overall cost, making it difficult to achieve widespread application in miniaturized, low-power wearable devices such as rings.

[0032] Therefore, the technical solution of this application is proposed, and the following is a detailed description of the technical solution of this application.

[0033] Figure 1 This is a flowchart illustrating a sleep apnea monitoring method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Acquire PPG signals of various color light sources collected by the wearable device at night.

[0034] Step 102: Obtain the user's blood oxygen signal based on the red PPG signal and infrared PPG signal from the PPG signals of multiple color light sources.

[0035] Step 103: Based on the local features of PPG signals and blood oxygen signals from multiple color light sources, and the feature matrix constructed by configuring weights for different signals, global features for characterizing the temporal sequence of changes in the user's sleep breathing events are obtained.

[0036] Step 104: Classify the continuous segment features of the first preset duration in the global features, and determine whether the sleep breathing of the user corresponding to the continuous segment feature has a hypoventilation event.

[0037] Step 105: When a hypoventilation event occurs, determine whether the blood oxygen signal shows an abnormal signal exceeding a second preset duration within the first preset duration.

[0038] Step 106: When an abnormal signal exceeding the second preset duration occurs, the hypoventilation event is confirmed to have actually occurred, indicating that the user experienced sleep apnea at night.

[0039] Specifically, sleep apnea syndrome (i.e., hypoventilation events) is a sleep disorder, which requires obtaining the user's physiological parameters during sleep. Therefore, the working time of the wearable device can be adjusted. For example, the working time of the wearable device can be set to 10 pm to 6 am the next day. At this time, physiological data can be collected from 10 pm to 6 am the next day. Of course, users can adjust the working time of the wearable device according to their own lifestyle habits. For example, users can set the working time directly through the controls on the wearable device, or they can set the working time of the wearable device through a smart terminal. The specific setting method is not limited here.

[0040] After setting the working time, the wearable device enters working mode when the working time arrives. Since the penetrating power of light emitted from different light sources is different, the physiological data that can be obtained will also be different. In order to obtain more accurate results and avoid inaccurate test results due to the use of a single light source, as well as the problem of inaccurate data obtained from a single light source due to the influence of ambient light, multiple colored light sources can be set on the side of the wearable device close to the user's skin. Multiple colored light sources can be lit sequentially at different times to obtain different physiological data. Since ambient light has different effects on different colored light sources, compared with a single light source, this design scheme can not only reduce the error rate of the acquired data, but also avoid the problem of inaccurate test results caused by an excessive weight of a single light source.

[0041] When sleep apnea syndrome occurs, the PPG signal will also change. In order to make the test results more accurate, blood oxygen content can be obtained by using different physiological information from various colored light sources. For example, green light with a wavelength of about 540nm is used to detect changes in blood volume in superficial blood vessels; red light with a wavelength of about 660nm is used to detect blood information in middle tissues; and infrared light with a wavelength of about 940nm is used to detect blood oxygen information in deep tissues. Therefore, the corresponding blood oxygen content can be determined by using red light PPG signals and infrared light PPG signals.

[0042] After obtaining PPG signals and blood oxygen content from multiple color light sources, this information can be processed using multi-layer 1D convolution + normalization + ReLU + maxpooling to obtain the local features of the PPG signals and blood oxygen content corresponding to each color light source. The local features include: details of the individual PPG pulse waveform, such as the steepness of the rising edge, the bluntness of the peak, and the local morphology of the dicrotic notch; instantaneous amplitude fluctuations, such as sudden decreases / increases in amplitude within a few beats (specifically, body motion, transient perfusion changes); short-term rhythm changes, such as slight acceleration / deceleration of the interbeat interval within a few seconds (local periodic changes); and short-term noise / artifact patterns, such as high-frequency jitter and sudden changes in local baseline.

[0043] At the same time, it is also necessary to assign weights to the PPG signals and blood oxygen content of each color light source. For example, the weight of each data point can be calculated using a global temporal attention (GTA) mechanism to more comprehensively capture the temporal dependencies in the time series, enabling the model to better understand the temporal patterns and dependencies in the time series data, thereby determining the importance of each data point.

[0044] After obtaining the weights, a feature matrix is ​​constructed based on the local features and their corresponding weights. Then, the global features used to characterize the temporal sequence of changes in the user's sleep breathing events are obtained through the constructed feature matrix. The global features include: the complete process of the breathing event, such as the entire pattern from "the onset of breathing restriction" to "recovery and rebound", rather than a single beat; the slow decline-recovery trajectory of SpO2, such as the fact that blood oxygen changes are usually lagging and have a long duration, requiring cross-time slice correlation; cross-modal time lag relationships, such as the correspondence of PPG abnormality first and SpO2 change later (or vice versa); and event density / periodic trends, such as whether the event recurs within a period of time (different from short-term single abnormality).

[0045] When confirming whether a hypopnea event has occurred, global features can be segmented according to a set step and time sliding window to obtain multiple continuous segment features. For each continuous segment feature, it needs to be input into a classifier for binary classification to obtain the probability of normal breathing and the probability of a hypopnea event. The one with the higher probability can be used as the result of binary classification. After determining that the continuous segment feature corresponds to a hypopnea event (i.e., the duration of the hypopnea event exceeds the first preset duration, such as 10 seconds), in order to confirm the accuracy of the hypopnea event, it is necessary to verify it from other dimensions. Since the occurrence of sleep apnea (i.e., the occurrence of a hypopnea event) is accompanied by an abnormal signal that lasts for more than the second preset duration, such as 10 seconds, the blood oxygen signal can be used to verify whether an abnormal signal exceeding the second preset duration occurs within the first preset duration to verify whether the user has sleep apnea at night. That is, the accuracy of the judgment result of whether a hypopnea event has occurred is verified by using abnormal signals.

[0046] In this application, multiple wavelength signals corresponding to various light sources are used to capture shallow and deep physiological data, and blood oxygenation signals are obtained from the acquired physiological data. Then, by analyzing the local features and different weights of various physiological data and blood oxygenation signals, global features characterizing the temporal sequence of changes in the user's sleep breathing events are obtained. By analyzing the features of continuous segments of the first preset duration in the global features, it is determined whether the user has experienced a hypopnea event, and whether there are any continuous abnormal signals after the hypopnea event. The accuracy of the hypopnea event is verified again by a secondary confirmation of abnormal signals, that is, verifying whether the user actually experienced sleep apnea at night. Since the above solution can be completed by wearable devices, the devices have the advantages of miniaturization, low power consumption and low cost. Furthermore, since the technical solution is obtained through comprehensive analysis of multiple data, it is beneficial to improve the accuracy of the results.

[0047] In a feasible implementation plan Figure 2 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment is shown below. Figure 2 As shown, step 102 is achieved through the following steps: Step 201: Based on the timestamps of the red PPG signal and the infrared PPG signal, obtain the red PPG signal segment and the infrared PPG signal segment of the third preset duration.

[0048] Step 202: For the red PPG signal segment and the infrared PPG signal segment, determine the peak value and valley value of the red PPG signal segment and the infrared PPG signal segment in each sliding window using preset sliding windows and steps.

[0049] Step 203: Calculate the blood oxygen value corresponding to each sliding window based on the peak and trough values ​​of the red PPG signal segment and the infrared PPG signal segment within each sliding window.

[0050] Step 204: Interpolate all the obtained blood oxygen values ​​according to the preset sampling frequency to obtain the blood oxygen signal.

[0051] Specifically, red light is used to detect blood information in the middle layer of tissue, while infrared light is used to detect blood oxygen information in the deep layer of tissue. By obtaining the red light PPG signal and the infrared light PPG signal at the same time, the blood oxygen value at that time can be obtained. In order to obtain the red light PPG signal and the infrared light PPG signal at the same time, the red light PPG signal and the infrared light PPG signal can be aligned according to the timestamp. Furthermore, in order to reduce the output processing load, a red light PPG signal segment and an infrared light PPG signal segment of a third preset duration can be extracted, such as obtaining the PPG signal segment from 0:00 to 3:00, or extracting a 1-minute PPG signal segment. Then, using preset sliding windows (e.g., 6 seconds) and steps (e.g., 2 seconds), the peak value and the valley value of the red light PPG signal segment and the infrared light PPG signal segment within each sliding window are determined. Then, the blood oxygen value corresponding to the sliding window is calculated. In order to make the frequency of the blood oxygen signal the same as that of the PPG signal segment, the blood oxygen value can be filled by interpolation to obtain the blood oxygen signal.

[0052] In a feasible implementation plan Figure 3 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment is shown below. Figure 3 As shown, the construction of the feature matrix in step 103, based on the local features of PPG signals and blood oxygen signals from multiple color light sources and the weights configured for different signals, is achieved through the following steps: Step 301: Perform maximum and minimum normalization processing on the PPG signals of various color light sources and the blood oxygen signal respectively to obtain the first value of the PPG signal of each color light source and the second value of the blood oxygen signal.

[0053] Step 302: Determine the first local feature of each first value and the second local feature of each second value based on the first value and the second value.

[0054] Step 303: Using the global attention mechanism, calculate the first weight of each of the first local features and the second weight of each of the second local features.

[0055] Step 304: Concatenate the product of the first local feature and the corresponding first weight, and the product of the second local feature and the second weight, to obtain the feature matrix.

[0056] Specifically, in order to avoid the excessive influence of data from a single dimension on the results, and to enable the calculation and comparison of multiple data sets, it is necessary to perform max-min normalization on the multiple data sets to ensure that the value ranges of the multiple data sets are the same. This can reduce the influence of data from a single dimension on the results, and the data sets can provide a basis for the calculation and comparison of multiple data sets together. Then, by using a global attention mechanism, the weight corresponding to each data set can be calculated, thereby more comprehensively capturing the time dependencies in the time series, enabling the model to better understand the time patterns and dependencies in the time series data.

[0057] In a feasible implementation, after obtaining the feature matrix, when obtaining global features to characterize the temporal sequence of changes in a user's sleep breathing events, the feature matrix is ​​used to extract features through a temporal feature model (such as a Transformer encoder) to obtain the global features.

[0058] In a feasible implementation plan Figure 4 A flowchart illustrating another sleep apnea monitoring method provided in this application embodiment is shown below. Figure 4 As shown, step 104 is achieved through the following steps: Step 401: Classify the continuous segment features using a classification model to determine the first probability of a normal breathing event and the second probability of a hypoventilation event.

[0059] Step 402: When the second probability is greater than the first probability, determine that the user corresponding to the continuous segment feature has a hypoventilation event during sleep.

[0060] Specifically, the classification model used in this application is pre-trained. The specific training process includes obtaining training samples (i.e., global features as training samples) through steps 101 to 103 and related content. Then, the classification model to be trained is trained through supervised training to obtain a stable classification model. After the classification model is trained, it can be used to infer and classify the user's sleep data. During the classification process, the probability of each result is given, and the time with the highest probability of occurrence is taken as the result of this classification.

[0061] It should be noted that data preprocessing may be involved during training, such as low-pass filtering of the original PPG. A more detailed training process can be set according to actual needs, and no specific limitations are made here.

[0062] In one feasible implementation, after determining that sleep apnea occurs within a first preset duration, the number of sleep apnea episodes per unit time can also be determined, and the number of episodes can be used as the sleep apnea-hypopnea index.

[0063] Specifically, the first preset duration is obtained according to the preset sliding window. Multiple first preset durations can be determined by setting the steps, thereby determining whether sleep apnea occurs within each first preset duration. This allows for the statistical analysis of the number of sleep apnea events per unit time (e.g., 1 minute or 1 hour), which is then used as the sleep-related apnea-hypopnea index (AHI) to provide data support for subsequent treatment or user health analysis.

[0064] In one feasible implementation, the wearable device is a wearable ring, a wearable headband, a wearable collar, a wearable armband, or a wearable legband.

[0065] It should be noted that the wearable devices involved in this application are devices that users can wear on their limbs. After the device is worn, it can acquire PPG signals from various light sources. Therefore, any wearable device that can perform the above-mentioned functions falls within the protection scope of this application.

[0066] In one feasible implementation, the wearable device includes: a microcontroller, a time-division driving circuit, multiple color light sources, and a photoelectric sensor. When acquiring PPG signals of multiple color light sources collected by the wearable device at night, the microcontroller first controls the time-division driving circuit to use time-division multiplexing to sequentially light up the corresponding color light sources according to a preset cyclic sequence. The photoelectric sensor acquires the current signal of the corresponding color light source, thus avoiding mutual interference between light sources. Then, according to a preset sampling frequency (e.g., 25Hz, to ensure that low-frequency changes related to breathing can be captured), the current signals of each color light source are digitally processed to obtain the original PPG signals of multiple color light sources. Finally, the original PPG signals of each color light source are low-pass filtered (e.g., passband 0.05-8Hz) to obtain the PPG signals of multiple color light sources.

[0067] Figure 5 This is a schematic diagram of the structure of a sleep apnea monitoring device provided in an embodiment of this application, as shown below. Figure 5 As shown, the device includes: Acquisition unit 51 is used to acquire PPG signals of multiple color light sources collected by the wearable device at night; The first determining unit 52 is used to obtain the user's blood oxygen signal based on the red light PPG signal and the infrared light PPG signal in the PPG signals of multiple color light sources; The second determining unit 53 is used to obtain global features that characterize the temporal sequence of changes in the user's sleep breathing events based on the local features of PPG signals and blood oxygen signals from multiple color light sources, as well as the feature matrix constructed by weights configured for different signals. The third determining unit 54 is used to classify the continuous segment features of the first preset duration in the global features and determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypoventilation event. The fourth determining unit 55 is used to determine whether the blood oxygen signal has an abnormal signal exceeding the second preset duration within the first preset duration when a hypoventilation event occurs. The fifth determining unit 56 is used to determine that the user has experienced sleep apnea at night when an abnormal signal exceeding the second preset duration occurs, thus confirming that the hypoventilation event has actually occurred.

[0068] In one feasible implementation, when the first determining unit obtains the user's blood oxygen signal based on the red PPG signal and the infrared PPG signal from multiple color light sources, it includes: Based on the timestamps of the red PPG signal and the infrared PPG signal, obtain the red PPG signal segment and the infrared PPG signal segment of the third preset duration. For red PPG signal segments and infrared PPG signal segments, the peak value and valley value of the red PPG signal segments and infrared PPG signal segments within each sliding window are determined using preset sliding windows and steps, respectively. Based on the peak and trough values ​​of the red PPG signal segment and the infrared PPG signal segment within each sliding window, calculate the blood oxygen value corresponding to each sliding window. The blood oxygen values ​​are interpolated according to a preset sampling frequency to obtain the blood oxygen signal.

[0069] In one feasible implementation, the second determining unit, when constructing a feature matrix based on the local features of PPG signals and blood oxygen signals from multiple color light sources, and the weights configured for different signals, includes: The PPG signals of various color light sources and the blood oxygen signal are subjected to maximum and minimum normalization processing respectively to obtain the first value of the PPG signal of each color light source and the second value of the blood oxygen signal. Based on the first value and the second value, determine the first local feature of each first value and the second local feature of each second value; Using a global attention mechanism, calculate the first weight of each first local feature and the second weight of each second local feature; The feature matrix is ​​obtained by concatenating the product of the first local feature and its corresponding first weight, and the product of the second local feature and its second weight.

[0070] In one feasible implementation, when the second determining unit obtains global features characterizing the temporal sequence of changes in a user's sleep breathing events, it includes: The global features are obtained by extracting features from the feature matrix using a time-series feature model.

[0071] In a feasible implementation, the third determining unit is used to classify continuous segment features of a first preset duration in the global features, and to determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypopnea event, including: The continuous segment features are classified using a classification model to determine the first probability of a normal breathing event and the second probability of a hypoventilation event. When the second probability is greater than the first probability, it is determined that the user corresponding to the continuous segment feature has a hypoventilation event during sleep.

[0072] In one feasible implementation, the device further includes: The sixth determining unit is used to determine the number of sleep apnea events occurring per unit time, and to use the number of events as the sleep apnea-hypopnea index.

[0073] In one feasible implementation, the wearable device includes: a microcontroller, a time-division driving circuit, multiple color light sources, and a photoelectric sensor; The acquisition unit is used to acquire PPG signals of multiple color light sources collected by the wearable device at night, including: The microcontroller controls the time-division drive circuit to use time-division multiplexing to sequentially light up the corresponding color light source according to the preset cycle sequence, and the photoelectric sensor obtains the current signal of the corresponding color light source. According to the preset sampling frequency, the current signals of each color light source are digitally processed to obtain the original PPG signals of multiple color light sources. The original PPG signals of each color light source are low-pass filtered to obtain PPG signals of multiple color light sources.

[0074] about Figure 5 For explanations of the principles behind the relevant content, please refer to the... Figures 1-4 The detailed explanation will not be repeated here.

[0075] Figure 6The diagram illustrates the structure of an electronic device according to an embodiment of this application, including a processor 601, a storage medium 602, and a bus 603. The storage medium 602 stores machine-readable instructions executable by the processor 601. When the electronic device runs a sleep apnea monitoring method, the processor 601 communicates with the storage medium 602 via the bus 603, and the processor 601 executes the machine-readable instructions to implement the steps of the aforementioned sleep apnea monitoring method.

[0076] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the sleep apnea monitoring method described above.

[0077] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0080] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0081] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0082] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for monitoring sleep apnea, characterized in that, The method includes: Acquire PPG signals of multiple color light sources collected by wearable devices at night; The user's blood oxygen signal is obtained from the red light PPG signal and the infrared light PPG signal in the PPG signals of multiple color light sources; Based on the local features of PPG signals and blood oxygen signals from multiple color light sources, and the feature matrix constructed with weights configured for different signals, global features for characterizing the temporal sequence of changes in sleep breathing events in users are obtained. Classify the continuous segment features of the first preset duration in the global features, and determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypoventilation event; When a hypoventilation event occurs, it is determined whether the blood oxygen signal exhibits an abnormal signal exceeding a second preset duration within the first preset duration; If an abnormal signal occurs for more than the second preset duration, the hypoventilation event is confirmed to have actually occurred, indicating that the user experienced sleep apnea at night.

2. The sleep apnea monitoring method as described in claim 1, characterized in that, The process of obtaining the user's blood oxygen signal based on the red PPG signal and infrared PPG signal from multiple color light sources includes: Based on the timestamps of the red PPG signal and the infrared PPG signal, obtain the red PPG signal segment and the infrared PPG signal segment of the third preset duration. For red PPG signal segments and infrared PPG signal segments, the peak value and valley value of the red PPG signal segments and infrared PPG signal segments within each sliding window are determined using preset sliding windows and steps, respectively. Based on the peak and trough values ​​of the red PPG signal segment and the infrared PPG signal segment within each sliding window, calculate the blood oxygen value corresponding to each sliding window. The blood oxygen values ​​are interpolated according to a preset sampling frequency to obtain the blood oxygen signal.

3. The sleep apnea monitoring method as described in claim 1, characterized in that, Based on the local features of PPG and blood oxygenation signals from multiple color light sources, and the weights assigned to different signals, a feature matrix is ​​constructed, including: The PPG signals of various color light sources and the blood oxygen signal are subjected to maximum and minimum normalization processing respectively to obtain the first value of the PPG signal of each color light source and the second value of the blood oxygen signal. Based on the first value and the second value, determine the first local feature of each first value and the second local feature of each second value; Using a global attention mechanism, calculate the first weight of each first local feature and the second weight of each second local feature; The feature matrix is ​​obtained by concatenating the product of the first local feature and its corresponding first weight, and the product of the second local feature and its second weight.

4. The sleep apnea monitoring method as described in claim 1, characterized in that, The obtained global features used to characterize the temporal sequence of changes in a user's sleep breathing events include: The global features are obtained by extracting features from the feature matrix using a time-series feature model.

5. The sleep apnea monitoring method as described in claim 1, characterized in that, The step of classifying continuous segment features of a first preset duration in the global features and determining whether the sleep breathing of the user corresponding to the continuous segment features has a hypopnea event includes: The continuous segment features are classified using a classification model to determine the first probability of a normal breathing event and the second probability of a hypoventilation event. When the second probability is greater than the first probability, it is determined that the user corresponding to the continuous segment feature has a hypoventilation event during sleep.

6. The sleep apnea monitoring method as described in claim 1, characterized in that, The method further includes: The number of sleep apnea episodes per unit time period is determined, and this number is used as the sleep apnea-hypopnea index.

7. The sleep apnea monitoring method as described in claim 1, characterized in that, The wearable device includes: a microcontroller, a time-division driving circuit, multiple color light sources, and a photoelectric sensor; acquiring the PPG signals of the multiple color light sources collected by the wearable device at night includes: The microcontroller controls the time-division drive circuit to use time-division multiplexing to sequentially light up the corresponding color light source according to the preset cycle sequence, and the photoelectric sensor obtains the current signal of the corresponding color light source. According to the preset sampling frequency, the current signals of each color light source are digitally processed to obtain the original PPG signals of multiple color light sources. The original PPG signals of each color light source are low-pass filtered to obtain PPG signals of multiple color light sources.

8. A sleep apnea monitoring device, characterized in that, The device includes: The acquisition unit is used to acquire PPG signals of various color light sources collected by the wearable device at night; The first determining unit is used to obtain the user's blood oxygen signal based on the red light PPG signal and the infrared light PPG signal from the PPG signals of multiple color light sources; The second determining unit is used to obtain global features that characterize the temporal sequence of changes in the user's sleep breathing events based on the local features of PPG signals and blood oxygen signals from multiple color light sources, as well as the feature matrix constructed by weights configured for different signals. The third determining unit is used to classify the continuous segment features of the first preset duration in the global features and determine whether the sleep breathing of the user corresponding to the continuous segment features has a hypoventilation event. The fourth determining unit is used to determine whether the blood oxygen signal has an abnormal signal exceeding a second preset duration within the first preset duration when a hypoventilation event occurs. The fifth determining unit is used to determine that the hypoventilation event has actually occurred when an abnormal signal exceeding the second preset duration is detected, thus confirming that the user has experienced sleep apnea at night.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 7.

Images