Device, apparatus and storage medium for sleep stage determination

Abstract

Device for determining sleep stages, characterized in that it comprises the following: a detection unit for capturing low-stress physiological signals from a target user; an extraction unit for extracting heartbeat features from the physiological signals based on the beat-to-beat interval in order to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; an analysis unit for analyzing heart rate variability based on the beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; a classification unit for classifying physiological characteristics using a pre-trained machine learning model to determine the sleep stage determination results of the target user. These physiological signals include an electrocardiographic signal and a ballistocardiogram signal; and wherein the extraction unit is used for the following purposes: extracting heartbeat features from the physiological signals based on the beat-to-beat interval by quadratic spline wavelet decomposition and R-wave peak detection to obtain a first sequence; extracting heartbeat features from the physiological signals based on the beat-to-beat interval by fitting the clustering template to obtain a second sequence; wherein the beat-to-beat interval sequence comprises the first sequence and the second sequence.

Description

TECHNICAL AREA

[0001] The present invention relates to the field of sleep monitoring technology, in particular a device, a apparatus and a storage medium for determining sleep stages. STATE OF THE ART

[0002] Conventional polysomnography devices face numerous challenges when used for sleep stage assessment in children, including complex operation, high cost, and low comfort and acceptance among children. Furthermore, the physiological signals of adults and children differ, and conventional sleep stage assessment methods used in adults may not be fully applicable to children, leading to poor sleep monitoring outcomes. Therefore, there is an urgent need for a sleep stage assessment method to improve the accuracy and acceptance of sleep stage assessment in children. CONTENT OF THE PRESENT INVENTION

[0003] To solve the technical problems described above, the embodiments of the present disclosure provide a device, a apparatus and a storage medium for determining sleep stages.

[0004] In a first aspect, an embodiment of the present disclosure provides a device for determining sleep stages, comprising: a detection unit for capturing low-stress physiological signals from a target user; an extraction unit for extracting heartbeat features from the physiological signals based on the beat-to-beat interval in order to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; an analysis unit for analyzing heart rate variability based on the beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; a classification unit for classifying physiological characteristics using a pre-trained machine learning model to determine the sleep stage determination results of the target user.

[0005] In a second aspect, an embodiment of the present disclosure provides an electronic device comprising: a storage facility; a processor; and a computer program; wherein the computer program is stored in memory and configured to be executed by the processor to perform a sleep stage determination procedure with the following steps: (a) Capturing low-stress physiological signals from a target user; (b) Extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; (c) Analyzing heart rate variability based on beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; (d) To implement classifying physiological features using a pre-trained machine learning model to determine the sleep stage determination results of the target user.

[0006] In a third aspect, an embodiment of the present disclosure provides a computer-readable storage medium in which a computer program is stored, wherein the above steps (a) - (d) are implemented when the computer program is executed by the processor.

[0007] The sleep stage determination performed using the device according to the invention comprises the following: capturing low-stress physiological signals from a target user; extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence, where the beat-to-beat interval refers to the time interval between two successive heartbeats; analyzing the heart rate variability based on the beat-to-beat interval sequence and extracting physiological features of the target user from multiple perspectives; classifying the physiological features using a pre-trained machine learning model to determine the sleep stage determination results of the target user.This fully takes into account the comfort of data collection in children, captures low-stress physiological signals, and performs sleep stage determination for children based on low-stress physiological signals, thereby improving the accuracy and acceptance of sleep stage determination for children and making it easier for the users concerned to understand the sleep status of children. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings are incorporated into the description and form an integral part of the description; they illustrate the implementation of the present invention and are used in conjunction with the description to explain the principles of the present invention.

[0009] To more clearly explain the technical solution in the embodiments of this disclosure or in the prior art, the drawings to be used in the explanation of the embodiment or for the prior art are briefly presented below. Obviously, a person skilled in the art in this field can derive other drawings from these drawings, provided that no creative work is undertaken. Fig. Figure 1 shows a flowchart for determining sleep stages; Fig. Figure 2 shows a flowchart of the beat-to-beat interval extraction from electrocardiogram signals; Fig. Figure 3(a) shows a result diagram of the feature point localization of the electrocardiogram signal and the beat-to-beat interval sequence; Fig.Figure 3(b) shows another result diagram of the feature point localization of the electrocardiogram signal and the beat-to-beat interval sequence; Fig. Figure 3(c) shows another result diagram of the feature point localization of the electrocardiogram signal and the beat-to-beat interval sequence; Fig. Figure 4 shows a flowchart of an algorithm for porous decomposition; Fig. Figure 5 shows a frequency-amplitude response curve diagram of an equivalent filter at a sampling frequency of 250 Hz; Fig. Figure 6 shows a schematic diagram of a 5-layer decomposition signal; Fig. Figure 7 shows a spectrum diagram of a 5-layer decomposition signal; Fig. Figure 8 shows a schematic diagram of a clustering template and an average template; Fig.Figure 9 shows a result diagram of the feature point localization of the ballistocardiogram signal and the beat-to-beat interval sequence; Fig. Figure 10 shows a flowchart of heart rate variability analysis and feature extraction; Fig. Figure 11 shows a schematic diagram of the structure of a device for determining sleep stages provided by an embodiment of the present disclosure; Fig. Figure 12 shows a schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0010] To better understand the aforementioned objectives, features, and advantages of this disclosure, the solution presented here will be explained in more detail below. It should be noted that the embodiments of this disclosure and the features in those embodiments can be combined without conflict.

[0011] Many specific details are described below to facilitate a complete understanding of the present disclosure, but the present disclosure can also be implemented in ways other than those described here; of course, the embodiments explained in the description represent only a part of the embodiments of the present disclosure, rather than all embodiments.

[0012] In particular, existing methods for determining sleep stages in adults are poorly suited for monitoring children's sleep. First, the electrocardiogram signal characteristics of children differ significantly from those of adults. Second, the classification performance of existing deep sleep and REM (rapid eye movement) stages is poor, and existing sleep stage determination methods are susceptible to interference under low signal-to-noise ratio conditions, which can lead to inaccurate stage determination results. Furthermore, conventional sleep monitoring devices are not suitable for daily monitoring of children because they are not user-friendly for children to wear, which hinders their widespread use among children.Therefore, existing methods for determining sleep stages in adults face significant challenges when extracting sleep signal features in children, particularly when extracting beat-to-beat intervals and performing HRV analyses under low-stress monitoring conditions, as errors can easily be introduced, which can affect the accuracy of sleep stage determination.

[0013] The following describes a method for determining sleep stages that extracts features from children's physiological signals, such as heart rate variability, and uses several conventional machine learning algorithms to classify physiological signals based on these extracted features, thereby achieving automated determination of children's sleep stages. This method can be applied to children's sleep monitoring devices to help caregivers understand children's sleep status.

[0014] The described method for determining sleep stages is applicable to scenarios for monitoring children's sleep. This method can be carried out by a sleep stage determination device according to the invention, which can be implemented in the form of software and / or hardware, and the system can be integrated into an electronic device. The electronic device can include, but is not limited to, mobile devices such as smartphones, laptops, digital radio receivers, personal digital assistants (PDAs), tablet PCs, portable multimedia players (PMPs), vehicle terminals (e.g., vehicle navigation terminals), wearable devices, etc., as well as fixed devices such as digital televisions, desktop computers, smart home devices, etc.

[0015] Fig. Figure 1 shows a procedure for determining sleep stages, including in particular the steps according to Fig. 1: S101: Detection of low-stress physiological signals from a target user.

[0016] It is understood that when capturing low-impact physiological signals from a target user, the target user may refer to users such as children for whom the use of conventional sleep monitoring devices is inconvenient. Specifically, low-impact physiological signals can be captured using monitoring devices suitable for children. "Suitable for children" refers to devices that fully consider factors such as monitoring comfort and acceptance by children. Low-impact physiological signals refer to physiological signals that have minimal impact on the subject during the capture process and are easier to monitor over the long term. These signals typically exhibit a high degree of comfort and low invasiveness, allowing them to be captured continuously in everyday settings without interfering with the subject's normal activities.

[0017] S102: Extracting heartbeat features from physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence.

[0018] The beat-to-beat interval refers to the time interval between two consecutive heartbeats.

[0019] It is understood that, based on S101, heartbeat features are extracted from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence. This reduces noise and disturbances in the physiological signals and ensures high robustness of the physiological signals. Particularly under conditions with a low signal-to-noise ratio, the features can also be extracted effectively. The beat-to-beat interval refers to the time interval between two successive heartbeats, typically measured in milliseconds (ms), and is frequently used in heart rate variability analysis. Heart rate variability (HRV) refers to the temporal variation of the intervals between heartbeats and serves as an important indicator for assessing the activity of the autonomic nervous system and cardiovascular health.

[0020] These physiological signals also include an electrocardiographic signal.

[0021] It is understood that an electrocardiogram (ECG) is a procedure for recording the electrical activity of the heart, in which electrical changes in the heart during each cardiac cycle are detected using electrodes placed on the skin surface.

[0022] Optionally, extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence can be achieved in particular through the following steps: Extracting heartbeat features from physiological signals based on the beat-to-beat interval using quadratic spline wavelet decomposition and R-wave peak detection to obtain a beat-to-beat interval sequence.

[0023] It is understood that the electrocardiogram signal is decomposed using quadratic spline wavelet decomposition, where the coefficients of the quadratic spline wavelet are simple. When decomposing the waveform of the electrocardiogram signal, the method not only filters out noise but also highlights the singular values ​​of the electrocardiogram signal. The singular values ​​reflect the main components of the electrocardiogram signal; for example, larger singular values ​​correspond to the main components of the electrocardiogram signal, while smaller singular values ​​typically correspond to noise or secondary components. Therefore, by using the wavelet decomposition method, the main components of the electrocardiogram signal can be extracted while simultaneously removing noise.Subsequently, by applying a squaring operation to enhance the peak advantages of the decomposed waveform, the R-wave peaks in the electrocardiogram signal can be identified based on their prominent peaks. This method is easy to computation and also exhibits high robustness. Finally, the beat-to-beat interval sequence of the electrocardiogram signal is determined based on the R-wave peaks.

[0024] Physiological signals also include a ballistocardiogram signal.

[0025] It is understood that the ballistocardiogram (BCG) signal is a technique that records cardiac activity by capturing the tiny mechanical vibrations the heart generates with each beat. The BCG signal reflects the changes in force and acceleration during the heart's pumping action, which can be detected with highly sensitive sensors.

[0026] Optionally, extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence can be achieved in particular through the following steps: Extracting heartbeat features from the physiological signals based on the beat-to-beat interval using quadratic spline wavelet decomposition and R-wave peak detection to obtain a first sequence; extracting heartbeat features from the physiological signals based on the beat-to-beat interval by fitting the clustering template to obtain a second sequence; wherein the beat-to-beat interval sequence comprises the first sequence and the second sequence.

[0027] Another possible approach is to extract heartbeat features from both the electrocardiogram (ECG) and ballistocardiogram (BCG) signals to obtain a beat-to-beat interval sequence. This means that the BCG sequence can be extracted from multiple physiological signals. Specifically, the first sequence is obtained by performing BCG extraction on the ECG signal. The BCG signal is less stable and more variable in shape compared to the ECG signal, and its main peak J-wave is easily influenced by other harmonic waves with a larger amplitude.Therefore, after preprocessing the acquired ballistocardiogram signal with a bandpass filter, an adaptive template-matching algorithm based on clustering is used to extract the beat-to-beat interval of the ballistocardiogram signal, thereby obtaining the second sequence. The clustering algorithm can be affinity propagation (AP). The beat-to-beat interval sequence comprises the first sequence and the second sequence, and the beat-to-beat interval sequence can be understood as the set of sequences obtained after the beat-to-beat interval extraction of each signal.

[0028] Optionally, the extraction of heartbeat features from the physiological signals based on the beat-to-beat interval can be achieved using quadratic spline wavelet decomposition and R-wave peak detection to obtain an initial sequence, in particular through the following steps: Using a moving average filter to preprocess the electrocardiogram signal and thereby eliminate high-frequency noise within the electrocardiogram signal; performing a quadratic spline wavelet decomposition on the preprocessed electrocardiogram signal to obtain a wavelet decomposition signal; squaring the wavelet decomposition signal to obtain a squared decomposition signal; locating a target peak point within the squared decomposition signal, where the target peak point is related to a maximum value that exceeds a preset threshold within the squared decomposition signal; searching for the R-wave peak point of the electrocardiogram signal within a preset window using the target peak point as a reference point;where the R-wave peak represents the maximum value among all target points contained in the electrocardiogram signal within the preset window range; calculating the time interval between two adjacent R-wave peaks to obtain the first sequence.

[0029] It is understood that a moving average filter is used to preprocess the electrocardiogram signal and remove high-frequency noise to reduce its impact on the feature point localization of the electrocardiogram signal waveform, thereby obtaining a preprocessed electrocardiogram signal. Specifically, a tenth-order moving average filter can be used for noise reduction, the formula of which is shown in equation (1). Subsequently, the preprocessed electrocardiogram signal is subjected to quadratic spline wavelet decomposition to obtain the wavelet decomposition signal. During wavelet decomposition, a scale with properties such as fewer coefficients, lower computational complexity, and good robustness can be selected.The wavelet decomposition signal is then squared to obtain the square wavelet decomposition signal (SWECG). This amplifies the peak amplitudes of the decomposition signal waveform. Furthermore, squaring increases the amplitude separation between the peak of the singular value and other disturbances. When subsequently applying threshold-based localization, this approach is less susceptible to disturbances and offers greater robustness compared to differential thresholding methods. The target peak point is then located within the squared decomposition signal. This target peak point corresponds to the maximum value that exceeds a preset threshold within the waveform of the squared decomposition signal. The setting of the preset threshold is related to the maximum amplitude value of the sampling point within the preset window.The size of the preset window is determined by the normal range of heart rate and sampling rate, typically covering 1–3 heartbeat cycles, with the dynamic updating of the preset threshold shown in equation (2). Subsequently, using the target peak as a reference, the R-wave peak of the electrocardiogram signal is located within the preset window. The R-wave peak is the maximum value point among all target value points contained in the electrocardiogram signal within the preset window range. The quadratic spline wavelet decomposition exhibits linear phase characteristics. There is a fixed time delay between the target peak of the decomposed signal waveform and the R-wave peak of the electrocardiogram signal waveform.Therefore, the size of the preset window can be adjusted based on the sampling rate, with the maximum value point of the electrocardiogram signal within the preset window representing the R-wave peak. The time interval between two adjacent R-wave peaks is then calculated to obtain the first sequence. Within the electrocardiogram signal, the R-wave peak marks the onset of ventricular depolarization, which signifies the beginning of cardiac contraction. The beat-to-beat interval typically refers to the time interval between two adjacent R-wave peaks, known as the RR interval. By detecting the R-wave peak in the electrocardiogram signal, a series of RR intervals is obtained, forming the beat-to-beat interval sequence. s(n)=110∑i=09y(n−i)

[0030] It says (n)for the value at the nth sampling point of the preprocessed electrocardiogram signal s, y for the original electrocardiogram signal and i for the index variable of the sliding window. Threshold=0.4*max(X), X=[x(n−400),…,x(n)]

[0031] Here, Threshold represents the preset threshold value, X represents the window size that SWECG of 400 sampling points, and x(n) represents the value of the SWECG at the nth sampling point.

[0032] Optionally, calculating the time interval between two adjacent R-wave peaks to obtain the first sequence can be achieved in particular by the following steps: Assess, based on the sampling rate corresponding to the preset window and the target user's heart rate range, whether there are any missing or multiple detections among the detected R-wave peaks; set the target user's historical interval sequence as the first sequence if there are missing or multiple detections; or calculate the time interval between two adjacent R-wave peaks to obtain the first sequence if there are no missing or multiple detections.

[0033] It is understood that if the electrocardiogram signal waveform is distorted, missing or multiple detections may occur when searching for R-wave peaks. To prevent this from affecting the beat-to-beat interval sequence, the presence of missing or multiple detections is assessed based on the sampling rate, which corresponds to the preset window, and the target user's heart rate range. If missing or multiple detections are present, the target user's historical interval sequence is set as the first sequence, meaning that historical values ​​are used as a substitute for RR interval values. If no missing or multiple detections are present, the time interval between two R-wave peaks is calculated to obtain the first sequence; that is, the RR interval value is calculated in real time.

[0034] For example, see Fig. 2. Fig.Figure 2 shows a flowchart of the beat-to-beat interval extraction from electrocardiogram signals; as in Fig.As shown in Figure 2, the procedure for extracting the beat-to-beat interval sequence comprises: 1) Acquiring a single-channel electrocardiogram signal, where the single-channel electrocardiogram signal refers to an ECG signal recorded using one channel (i.e., one pair of electrodes) and subjected to sliding average filtering. Single-channel ECG devices are typically compact, lightweight, portable, and user-friendly, making them suitable for home health monitoring and personal health management; 2) Performing sliding average filtering on the single-channel electrocardiogram signal; 3) Performing quadratic spline wavelet decomposition on the filtered electrocardiogram signal; 4) Squaring the decomposed waveform; 5) Locating the peaks within the squared decomposition signal.6) Assess whether the peak value in the squared decomposition waveform exceeds a preset amplitude threshold; 7) Determine the target peak on the squared decomposition waveform if the peak value exceeds the preset amplitude threshold; 8) Locate the maximum point on the ECG within a preset window using the target peak as a reference and determine this as the R-wave peak point; 9) Assess whether there are any missing or multiple detections; 10) Calculate the RR interval based on the R-wave peak point if there are no missing or multiple detections to obtain a beat-to-beat interval sequence;10) Determining the historical interval as the RR interval if the peak value is smaller than the preset amplitude threshold or if there are missing or multiple detections. For a detailed explanation of steps 1) to 10), which are affected by the beat-to-beat interval sequence extraction process, please refer to the above examples, which will not be discussed further here.

[0035] For example, reference is made to a result diagram of the feature point localization of the electrocardiogram signal and the beat-to-beat interval sequence, in particular the diagram of the wavelet decomposition-based localization of the R-wave peak and the beat-to-beat interval sequence, including the result diagram under normal conditions, as in Fig. 3(a) shows the result diagram under conditions with significant disturbances, as in Fig.3(b) is shown, and the result diagram under conditions with pronounced noise, as in Fig. 3(c) is shown. The outcome plot of each scenario displays the waveforms for the ECG, SWECG, and beat-to-beat interval sequence. When significant ECG disturbances occur, wavelet-based R-wave localization tends to be misaligned. As shown in Fig. As shown in Figure 3(b), significant disturbances in ECG within a range of 12,000 to 8,000 sampling points result in the SWECG subsequently obtained by wavelet decomposition and squaring operations having misaligned peaks. However, error detection prevents significant disturbances from affecting the beat-to-beat interval sequence. Even with significant high-frequency ECG noise, as in Fig.As shown in Figure 3(c), SWECG retains pronounced peak advantages and a high signal-to-noise ratio. Exploiting this advantage, the beat-to-beat interval sequence of ECG with greater noise than the beat-to-beat interval sequence of ECG under normal conditions can precisely localize each cardiac cycle.

[0036] Optionally, performing a quadratic spline wavelet decomposition on the preprocessed electrocardiogram signal to obtain a wavelet decomposition signal can be achieved in particular by the following steps: Construct an equivalent filter corresponding to wavelets at different scales, based on low-pass and high-pass filters; determine the target scale based on the frequency-amplitude response curves of the equivalent filter at multiple scales under a preset sampling frequency and the frequency range of the QRS complex; perform wavelet decomposition on the preprocessed electrocardiogram signal to obtain the wavelet decomposition signal at the target scale.

[0037] It is understood that the quadratic spline wavelet template is simple and possesses inverse properties, thus exhibiting good performance in detecting signal singular values. The Fourier transforms of the wavelet function and the corresponding high-pass and low-pass filters are shown in equations (3) to (5). ψ(ω)=jω(sinω4ω4)4 H(ω)=18e−jω+38+38ejω+18ej2ω G(ω)=−2+2ejw

[0038] Here, Ψ(ω) represents the wavelet function, H(ω) the low-pass filter, G(ω) the high-pass filter, ω the angular frequency, and j the imaginary unit that j 2 =-1 is fulfilled.

[0039] It follows from equations (4) and (5) that the high-pass and low-pass filters corresponding to the quadratic spline wavelet function have low complexity, which simplifies computation. This approach can save operating time when preprocessing a large number of electrocardiogram signals.

[0040] It is understood that wavelet processing introduces problems with the reduced sampling rate in the layered decomposition process. To ensure that the decomposed signal maintains the same sampling rate as the original signal for subsequent feature point localization, a porous algorithm is used. This involves 2 j- One zero point is inserted into the filter to ensure that the length of the wavelet-decomposed signal waveform matches the original signal, where j represents the wavelet decomposition scale. For example, see Fig. 4, shows Fig. 4 A flowchart of a porous decomposition algorithm provided by an embodiment of the present disclosure. This porous decomposition algorithm decomposes a signal into different resolution layers by means of a multiscale analysis, each layer representing different frequency components of the signal. Fig. Figure 4 illustrates a process in which the high-pass filter (G) and the low-pass filter (H) each decompose the signal (X) into several resolution layers, which is not explained in more detail here.

[0041] It is understood that the determination of the scales in wavelet decomposition is explained as follows: the expression of the quadratic spline wavelet scaling function in the frequency domain is given in equation (6), and its relationship to the low-pass filter is shown in equation (7). The relationship between the wavelet function, the high-pass filter, and the scaling function in the frequency domain is given by equation (8). The equivalent filter corresponding to the wavelet at different scales is shown in equation (9); in particular, the equivalent filter in equation (9) can be derived from equations (4) and (5) above. γ(ω)=(sin(ω / 2)ω / 2)3e(−jω / 2) γ(2ω)=H(ω)γ(ω) φ(2ω)=G(ω)γ(ω) Q(ω)={G(ω)j=1G(2ω)H(ω)j=2G(2j−1ω)H(2j−2ω)… …H(ω)j>2

[0042] Here, γ(ω) represents the scaling function, γ(2ω) characterizes the relationship between the low-pass filter and the scaling function, and φ(2ω) characterizes the relationship between the high-pass filter and the scaling function.

[0043] For example, see Fig. 5. Fig. Figure 5 shows a frequency-amplitude response curve diagram of an equivalent filter at a sampling frequency of 250 Hz, that is, when the sampling rate is 250 Hz, the following values ​​are shown in the diagram: Fig. 5 the frequency response curves of the equivalent filter for the quadratic spline wavelet decomposition are shown for the first five scales, where the five scales are j=1, 2....5.

[0044] For example, see Fig. 6. Fig. Figure 6 shows a schematic diagram of a 5-layer decomposition signal, i.e., a schematic diagram obtained by wavelet decomposition at scales j = 1, 2, ..., 5. As can be seen from the in Fig.As can be seen in the schematic diagram of the multiscale decomposition signal shown in Figure 6, the high-frequency information is concentrated mainly in the lower scale levels, while the low-frequency information is concentrated mainly in the higher scale levels.

[0045] For example, see Fig. 7. Fig. Figure 7 shows a spectrum diagram of a 5-layer decomposition signal. Fig. Figure 7 shows the ECG and the spectral diagrams of the decomposed waveforms in five scales. Since the frequency range of the QRS complex is typically between 5 Hz and 22 Hz, it can be determined from Fig.As can be seen in Figure 7, the energy of the QRS complex is concentrated on scales 3 and 4. Considering both the time complexity and the robustness of the algorithm, analyzing the decomposed waveform on scale 3 is therefore the optimal solution. Scale 3 is thus defined as the target scale, and its selection for wavelet decomposition analysis offers advantages such as fewer coefficients, lower computational complexity, and good robustness. The QRS complex reflects changes in the depolarization potential and timing within the left and right ventricles. The initial downward wave is the Q wave, followed by the upward wave R, with the subsequent downward wave being the S wave.

[0046] It is understood that after determining the target scale, a wavelet decomposition is performed on the preprocessed electrocardiogram signal to obtain a wavelet decomposition signal of scale 3. The wavelet decomposition formula for scale 3 is shown in equation (10). ws(n)=((s(n+10)+3*s(n+9)+6*s(n+8)+10*s(n+7)+11*s(n+6)+9*s(n+5)+4*s(n +4)−4*s(n+3)−9*s(n+2)−11*s(n+1)−10*s(n)−6*s(n−1)−3*s(n−2)−s(n−3)) / 32

[0047] Here, s represents the filtered signal, ws the wavelet decomposition signal of scale 3, and n the current position.

[0048] Optionally, extracting heartbeat features from the physiological signals based on the beat-to-beat interval can be achieved by fitting the clustering template to obtain a second sequence, in particular by the following steps: Using a bandpass filter to filter the ballistocardiogram signal in order to obtain the subtle oscillation signal associated with the heartbeats within the ballistocardiogram signal; extracting all maximum points from the subtle oscillation signal and capturing the data samples within a predetermined time period from the subtle oscillation signal with these maximum points as the center; clustering the data samples using the affinity spread clustering algorithm to obtain a clustering template; extracting the heartbeat features from the subtle oscillation signal based on the beat-to-beat interval according to the clustering template to obtain the second sequence.

[0049] It is understood that the extraction from the ballistocardiogram signal based on the beat-to-beat interval is achieved in particular through the following steps: A bandpass filter is used to filter the ballistocardiogram signal (BCG) in order to obtain the subtle vibrational component associated with the heartbeat movements within the ballistocardiogram signal. The bandpass filter can be a Butterworth bandpass filter of 8–24 Hz, and the ballistocardiogram signal can be acquired via a piezoelectric ceramic sensor. Subsequently, all maximum points within the filtered ballistocardiogram signal (subtle vibrational signal) are extracted, and with the maximum point as the center, a signal of a predetermined time interval is extracted from the subtle vibrational signal as a sample b. A dataset is created from all extracted samples, e.g., dataset B = {b0, b1, ..., b n}, where n is the number of samples and b_i represents the signal segment extracted at the i-th peak value. The predetermined time interval can be approximately 0.4 seconds. After obtaining the sample set, the affinity spread clustering algorithm is used to perform clustering on sample set B, yielding at least one clustering template. Subsequently, the means of each clustering template within at least one clustering template are averaged. The cluster with the highest average value, or the one within the maximum range of values, is selected as the target cluster. All target clusters are then averaged to obtain the BCG mean template.Subsequently, the heartbeat features are extracted from the subtle oscillation signal based on the beat-to-beat interval in accordance with the BCG average template to obtain a second sequence. This algorithm for fitting the clustering template ensures accuracy in identifying key components of the ballistocardiogram signal and forms the basis for subsequent heart rate variability calculations.

[0050] For example, see Fig. 8. Fig. Figure 8 shows a schematic diagram of a clustering template and an average template. The affinity-spreading clustering algorithm is used to perform clustering on sample set B in order to create a BCG clustering template according to... Fig.8. The means of each clustering template are averaged, with the target cluster being the one with the maximum average value. Then, all target clusters are averaged to obtain the BCG average template according to... Fig. to receive 8.

[0051] For example, see Fig. 9. Fig. Figure 9 shows a result plot of the feature point localization of the ballistocardiogram signal and the beat-to-beat interval sequence. The BCG average template determined above is used to test the beat-to-beat interval extraction algorithm of the ballistocardiogram signal. The raw ballistocardiogram signal acquired by the piezoelectric ceramic front end, the filtered heartbeat signal, the correlation coefficient, and the extracted cardiac cycle (the second sequence) are shown in Figure 9. Fig. 9 shown.

[0052] S103: Analyzing heart rate variability based on beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives.

[0053] It is understood that, based on S102, a heart rate variability (HRV) analysis is performed using beat-to-beat interval sequences, extracting physiological characteristics from multiple perspectives. Heart rate variability encompasses continuous beat-to-beat interval sequences in cardiac activity and describes minute fluctuations within the cardiac cycle. It reflects the degree of interaction between the regulation of the heart by the sympathetic and parasympathetic nervous systems and summarizes physiological information about the human cardiovascular system. HRV exhibits different patterns in various sleep stages, which is why HRV and its properties are a frequent focus of research on sleep stage determination with low physiological stress.

[0054] Optionally, analyzing heart rate variability based on the beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives can be achieved, in particular, through the following steps: Performing a re-sample on the beat-to-beat interval sequence to obtain a uniformly sampled sequence; removing the DC signal from the uniformly sampled sequence to obtain a DC removal sequence; analyzing heart rate variability from multiple perspectives based on the DC removal sequence to extract the target user's physiological characteristics, including linear perspectives, nonlinear perspectives, and frequency ranges; standardizing the physiological characteristics to obtain physiological characteristics that conform to a standard normal distribution.

[0055] It is understood that the beat-to-beat interval sequence represents an unevenly sampled time series. The horizontal axis of the signal represents the cumulative sum of the historical interval sequences at that point, while the vertical axis represents the interval value at that point. The interval value is defined as the difference in the number of sampling points of the two R-wave peaks. Frequency domain analysis of the beat-to-beat interval sequence assumes that the signal is an evenly sampled sequence. Therefore, re-sampling the beat-to-beat interval sequence is necessary to obtain an evenly sampled sequence. The normal range of the beat-to-beat interval in healthy individuals is typically between 300 and 2000 milliseconds, with a maximum frequency of 0.5 Hz. Therefore, a re-sampling frequency of 2 Hz can be set to satisfy the Shannon-Nyquist sampling theorem.Subsequently, cubic spline interpolation is used to resample the beat-to-beat interval sequence, resulting in a smooth curve at the target sampling rate. It is understood that if the beat-to-beat interval sequence includes sequences extracted from multiple types of physiological signals based on the beat-to-beat interval, the heart rate variability analysis is performed on each type of physiological signal sequence separately. For example, if the first and second sequences are extracted as described above, the heart rate variability analysis is performed on the first sequence and the second sequence respectively, without any prescribed order of analysis.After resampling the signal, the uniformly sampled sequence undergoes DC removal. Specifically, the resampled beat-to-beat interval sequence (uniformly sampled sequence) contains DC components. These DC signals must be eliminated before power spectral analysis to prevent other AC frequency components from being missed during frequency domain analysis due to excessive DC components. After DC removal, HRV analysis is performed from three perspectives: time domain, frequency domain, and nonlinearity, to extract physiological features. The relevant physiological features are listed in Table 1. Following feature extraction, the physiological features are standardized to obtain physiological features that follow a standardized normal distribution. In particular, physiological features include multidimensional features.Given the different attributes and numerical magnitudes of the features in each dimension, significant feature differences can occur. When multidimensional features differ by several orders of magnitude, the learner may overfocus on the higher-order-of-magnitude features and thus fail to learn from the features with smaller differences. Therefore, to ensure numerical comparability between different dimensions and improve the accuracy of the classifier, standardization of the multidimensional features is necessary. For example, the Z-score method (standardized score) can be used, where the data are standardized by subtracting the mean and dividing by the standard deviation. The processed data then conform to a standard normal distribution.The Z-score is a statistical concept that measures the number of standard deviations by which a data point deviates from the mean of the data set. Table 1: feature Description of the parameters Time domain analysis Mean, standard deviation, median Mean, standard deviation and median of the interval percentile Cumulative percentile in ascending order of the values ​​of the interval, reflecting the data distribution between maximum and minimum values. quartile deviation Difference between the 75th and 25th percentiles, indicating the data dispersion. coefficient of variation The ratio of standard deviation to mean compares the degree of variation between individuals with significantly different mean values. RMSSD Square mean of the differences between adjacent intervals NN50, PNN50, NN20, PNN20 Number of adjacent intervals with differences greater than 50 ms or 20 ms; NN50 and NN20 indicate the percentage of the total number of intervals. Triangle Index Height of the total number of intervals within the interval histogram Average heart rate Average heart rate from beat to beat Frequency domain analysis Energy values, maximum energy value, percentage of total energy Total energy values ​​for the frequency bands Ultra-Low Frequency (ULF), Very Low Frequency (VLF), Low Frequency (LF) and High Frequency (HF), maximum energy peak, percentage of the energy of each band in relation to the total energy (TP) Normalized energy LF n = LF / (TP - VLF)HF n = HF / (TP - VLF) Energy ratio LF / HF, VLF / (LF + HF) Nonlinear analysis Sample entropy Complexity of time series measurements Cardiac sympathetic index Ratio of longitudinal to transverse length in the Lorenz diagram Cardiac vagal index Logarithm (base 10) of the product of longitudinal and transverse length in the Lorenz diagram Poincare scatter plot index Standard deviation of the Poincaré diagram projection, standard deviation of the projections onto lines perpendicular to the isoclines

[0056] For example, see Fig. 10. Fig. Figure 10 shows a flowchart for heart rate variability analysis and feature extraction. For the beat-to-beat interval sequence of a single physiological signal, heart rate variability analysis and feature extraction are performed by the following steps, as shown in Fig.Figure 10 illustrates: 1) Acquiring the beat-to-beat interval sequence; 2) Re-sample the beat-to-beat interval sequence; 3) Performing DC removal on the re-sampled beat-to-beat interval sequence; 4) Performing feature extraction on the beat-to-beat interval sequence after DC removal; 5) Obtaining time-domain features, frequency-domain features, and nonlinear features; 6) Standardizing the acquired multiscale features. It is understood that the detailed descriptions of steps 1) through 6) refer to the embodiments mentioned above, which are not further explained here.

[0057] S104: Classifying physiological features using a pre-trained machine learning model to determine the sleep stage determination results of the target user.

[0058] It is understood that, based on S103, after performing a heart rate variability analysis and feature extraction on the beat-to-beat interval sequence, the physiological features are classified using a pre-trained machine learning model to determine the target user's sleep stage determination results. Typically, sleep stage determination is divided into five main stages: N1 (non-REM sleep stage 1), N2 (non-REM sleep stage 2), N3 (non-REM sleep stage 3), and REM (rapid eye movement). Each stage exhibits distinct physiological features. The specific criteria for delineating these stages are not fixed and can be adapted according to the user's requirements. The machine learning model can employ at least one of the following methods: random forest, multi-layered perceptron, logistic regression, Adaboost, or support vector machine.Several learning models can be used to train sleep stage determination based on heart rate variability features extracted from the single-channel electrocardiogram (ECG) and ballistocardiogram (BCG) signals, and the results are analyzed in both non-independent and independent subject modes. In non-independent subject mode, data from all subjects are read directly and then proportionally randomized, with one portion designated as the training set and the other as the test set. However, when using non-independent subject mode for data allocation, the data within the training and test sets may originate from the same individual. In contrast, independent subject mode first divides the subject population into training and test cohorts and then reads the data from each cohort separately to form the training and test sets.Thus, in independent subject mode, there is no overlap of subjects between the data in the training set and the test set, which improves training accuracy. Specifically, multiple models can be trained, and the model with better accuracy is selected as the machine learning model for predicting sleep stage determination.

[0059] A suitable method for sleep stage determination captures a user's physiological signals, extracts the beat-to-beat interval sequence from these signals using quadratic spline wavelet decomposition, and performs heart rate variability (HRV) analysis and physiological feature extraction. Furthermore, during the feature extraction phase, HRV is analyzed from multiple perspectives, including time domain, frequency domain, and nonlinearity, to extract physiological features indicative of sleep. Subsequently, several traditional machine learning algorithms (such as random forests, support vector machines, and logistic regression) are employed to classify children's sleep stages based on the extracted physiological features.This method is adapted to the physiological characteristics of children, significantly improving the accuracy of sleep stage determination and effectively processing physiological signals under low signal-to-noise ratio conditions. Furthermore, compared to conventional polysomnography methods, this sleep stage determination technique can be integrated into wearable monitoring devices, making it suitable for routine sleep monitoring in children. Additionally, the robust signal processing and sleep stage determination technique can provide more reliable sleep data, offering relevant users a scientific reference and aiding in the early detection and intervention of sleep problems in children.

[0060] Fig.Figure 11 shows a schematic diagram of the structure of a sleep stage determination device according to the invention, provided by an embodiment of the present disclosure. The sleep stage determination device provided by the embodiment of the present disclosure can perform the sleep stage determination method described above using the provided processing sequence. As in Fig. As shown in 11, the device 1100 comprises the following: a 1101 acquisition unit for capturing low-stress physiological signals from a target user; an extraction unit 1102 for extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence; wherein the beat-to-beat interval refers to the time interval between two successive heartbeats; an analysis unit 1103 for analyzing heart rate variability based on the beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; a classification unit 1104 for classifying physiological characteristics using a pre-trained machine learning model to determine the sleep stage determination results of the target user.

[0061] These physiological signals optionally include an electrocardiographic signal and a ballistocardiogram signal.

[0062] Optionally, the extraction unit 1102 is used for the following purposes: Extracting heartbeat features from physiological signals based on the beat-to-beat interval using quadratic spline wavelet decomposition and R-wave peak detection to obtain an initial sequence; Extracting heartbeat features from physiological signals based on the beat-to-beat interval by fitting the clustering template to obtain a second sequence; where the beat-to-beat interval sequence includes the first sequence and the second sequence.

[0063] Optionally, the extraction unit 1102 is used for the following purposes: Using a sliding average filter to preprocess the electrocardiogram signal and thereby eliminate high-frequency noise within the electrocardiogram signal; Performing a quadratic spline wavelet decomposition on the preprocessed electrocardiogram signal to obtain a wavelet decomposition signal; Squaring the wavelet decomposition signal to obtain a squared decomposition signal; Locating a target peak point within the squared decomposition signal; wherein the target peak point refers to a maximum value that exceeds a preset threshold value within the squared decomposition signal;

[0064] Searching for the R-wave peak of the electrocardiogram signal within a preset window using the target peak as a reference point; wherein the R-wave peak represents the maximum value among all target points contained in the electrocardiogram signal within the preset window range;

[0065] Calculating the time interval between two adjacent R-wave peaks to obtain the first sequence.

[0066] Optionally, the extraction unit 1102 is used for the following purposes: Assess, based on the sampling rate corresponding to the preset window and the target user's heart rate range, whether there are any missing or multiple detections among the detected R-wave peaks; Setting the target user's historical interval sequence as the first sequence if there are missing or multiple detections; or calculating the time interval between two adjacent R-wave peaks to obtain the first sequence if there are no missing or multiple detections.

[0067] Optionally, the extraction unit 1102 is used for the following purposes: Constructing an equivalent filter that corresponds to wavelets at different scales, based on low-pass and high-pass filters; Determining the target scale based on the frequency-amplitude response curves of the equivalent filter in several scales under a preset sampling frequency and the frequency range of the QRS complex; Perform wavelet decomposition on the preprocessed electrocardiogram signal to obtain the wavelet decomposition signal on the target scale.

[0068] Optionally, the extraction unit 1102 is used for the following purposes: Using a bandpass filter to filter the ballistocardiogram signal in order to obtain the subtle oscillation signal associated with the heartbeats within the ballistocardiogram signal; Extracting all maximum points from the subtle vibration signal and capturing the data samples within a predetermined time period from the subtle vibration signal with these maximum points as the center; Using the affinity propagation clustering algorithm to cluster the data samples, resulting in a clustering template; Performing a heartbeat feature extraction on the subtle oscillation signal based on the beat-to-beat interval in accordance with the clustering template to obtain a second sequence.

[0069] Optionally, the analysis unit 1103 is used for the following purposes: Performing a re-scan of the beat-to-beat interval sequence to obtain a uniformly scanned sequence; Removing the DC signal from the uniformly sampled sequence to obtain a DC removal sequence; Analyzing heart rate variability from multiple perspectives based on the DC distance sequence to extract the physiological characteristics of the target user; where these perspectives include linear perspectives, nonlinear perspectives, and frequency ranges; Standardizing physiological characteristics to obtain physiological characteristics that correspond to a standard normal distribution.

[0070] The device for determining sleep stages in the exemplary embodiment according to Fig. 11 can be used to implement the technical solution in the above procedure and exhibits analogous functional principles and technical effects, which will not be explained again here.

[0071] Fig. Figure 12 shows a schematic diagram of the structure of an electronic device according to the invention provided by an embodiment of the present disclosure. With particular reference to Fig. Figure 12 shows a schematic diagram of the structure of an electronic device 1200 suitable for implementing sleep stage determination. The electronic device 1200 in the embodiment of the present disclosure may include, but is not limited to: mobile devices such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), tablets (PADs), PMPs (Portable Multimedia Players), vehicle terminals (e.g., vehicle navigation devices), and portable electronic devices, as well as fixed devices such as digital televisions, desktop computers, and smart home devices, etc. The Fig. The electronic device shown in Figure 12 is merely an example and should not impose any limitations on the functionality or scope of the embodiments described in this disclosure.

[0072] As in Fig.As shown in Figure 12, the electronic device 1200 can comprise a processing unit (e.g., a main processor, a graphics processing unit, etc.) 1201, which performs various suitable actions and processing operations to implement the sleep-stage determination method in the embodiment of the present disclosure, based on a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage device 1208 into a random-access memory (RAM) 1203. The RAM 1203 also stores various programs and data required for the operation of the electronic device 1200. The processing unit 1201, the ROM 1202, and the RAM 1203 are interconnected via a bus 1204. An input / output interface (I / O interface) 1205 is also connected to the bus 1204.

[0073] Typically, the following devices can be connected to the I / O interface 1205: an input device 1206, including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; an output device 1207, including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; a storage device 1208, including, for example, magnetic tapes, hard disks, etc.; and a communication device 1209. The communication device 1209 can enable the electronic device 1200 to communicate wirelessly or via a wired connection with other devices for data exchange. Although Fig. Figure 12 shows the electronic device 1200 with various devices; it should be clear that the implementation or inclusion of all the devices shown is not required. Alternative implementations may include more or fewer devices.

[0074] In particular, according to the embodiments of this disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, one embodiment of this disclosure comprises a computer program product that includes a computer program stored on a non-transitory, computer-readable medium, wherein the computer program includes program code for executing the procedure shown in the flowchart, thereby implementing the sleep-stage determination procedure described above. In such embodiments, the computer program can be downloaded and installed from a network via the communication device 1209, installed from a storage device 1208, or installed from a ROM 1202.When executed by a processing device 1201, the computer program performs the above-mentioned functions, which are defined in the procedures described above.

[0075] It should be noted that the computer-readable medium described herein may be a computer-readable signaling medium, a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or apparatus, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random-access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.In the present disclosure, a computer-readable storage medium can be any material medium that contains or stores a program, wherein the program can be used by or in conjunction with a command execution system, device, or apparatus. In the present disclosure, a computer-readable signaling medium can comprise a data signal that is contained in the baseband or transmitted as part of a carrier and carries computer-readable program code. Such transmitted data signals can take various forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof.A computer-readable signaling medium can also be any computer-readable medium other than a computer-readable storage medium that is capable of transmitting, disseminating, or disseminating a program for use by or in conjunction with a command-execution system, device, or apparatus. The program code contained on a computer-readable medium can be transmitted via any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0076] In some implementations, clients and servers can communicate using any currently known or future developed network protocols, such as HTTP (Hypertext Transfer Protocol), and be connected to each other via digital data communication in any form or on any media (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the internet (e.g., the World Wide Web), and end-to-end networks (e.g., ad-hoc end-to-end networks), as well as any other currently known or future developed networks.

[0077] The above computer-readable medium may be integrated into the above electronic device; alternatively, it may exist separately and not be built into the electronic device.

[0078] Optionally, when one or more of the above programs are executed by the electronic device, it can also perform other steps as described in the preceding examples.

[0079] The computer program code for performing the operations of this disclosure may be written in one or more programming languages ​​or combinations thereof, including, but not limited to, object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In remote computer scenarios, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or establish a connection to an external computer (e.g., a server).(using an internet service provider to connect via the internet).

[0080] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of the systems, methods, and computer program products according to the embodiments described in this disclosure. In this context, each block in the flowchart or block diagram can represent a module, a program segment, or a portion of code, wherein the module, program segment, or portion of code contains one or more executable instructions for implementing a defined logical function. It should also be noted that in some alternative implementations, the functions specified in the blocks may occur in a different order than that shown in the accompanying drawings. For example, two consecutive blocks may be executed essentially in parallel, or, depending on the function in question, they may sometimes be executed in reverse order.It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a special hardware-based system that performs the specified function or action, or it can be implemented by a combination of special hardware and computer instructions.

[0081] The units involved in the embodiments described in this disclosure can be implemented in software or in hardware. In some cases, the designation of the units does not constitute a limitation of the units themselves.

[0082] The functions described here can be performed, at least partially, by one or more hardware logic components. Examples of non-restrictive types of hardware logic components include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0083] In the context of this disclosure, a machine-readable medium can be a physical medium capable of containing or storing a program for use by or in conjunction with a command-execution system, device, or apparatus. A machine-readable medium can be a machine-readable signaling medium or a machine-readable storage medium. Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof.More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0084] It should be noted that relational terms such as "first," "second," etc., in the description serve only to distinguish one object or action from another, without necessarily requiring or implying that any such actual relationship or sequence exists between the objects or actions. Furthermore, terms such as "include," "exhibit," or other variants cover non-exclusive exhibiting, so that a process, procedure, object, or device encompassing a series of elements includes both such elements and other elements not explicitly listed or inherent to that process, procedure, object, or gateway. If no further restrictions apply, a process or device described as "encompassing a series of elements" will...The defined element does not exclude the possibility that other identical elements exist within a process, procedure, object, or gateway that encompasses the element.

[0085] The above content represents only one specific embodiment of the present disclosure, enabling the person skilled in the art to understand and implement the present disclosure. The various modifications of the embodiments are obvious to the person skilled in the art, and the general principles defined in the description can be realized in other embodiments without deviating from the intent or scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments described above, but rather the present invention encompasses the broadest range that is consistent with the disclosed principles and creative points of the present description.

Claims

[1] Device for determining sleep stages, characterized by that it includes the following: a detection unit for capturing low-stress physiological signals from a target user; an extraction unit for extracting heartbeat features from the physiological signals based on the beat-to-beat interval in order to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; an analysis unit for analyzing heart rate variability based on the beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; a classification unit for classifying physiological characteristics using a pre-trained machine learning model to determine the sleep stage determination results of the target user. These physiological signals include an electrocardiographic signal and a ballistocardiogram signal; and wherein the extraction unit is used for the following purposes: extracting heartbeat features from the physiological signals based on the beat-to-beat interval by quadratic spline wavelet decomposition and R-wave peak detection to obtain a first sequence; extracting heartbeat features from the physiological signals based on the beat-to-beat interval by fitting the clustering template to obtain a second sequence; wherein the beat-to-beat interval sequence comprises the first sequence and the second sequence. [2] Electronic device, characterized by , that it includes: a storage facility; a processor; and a computer program; wherein the computer program is stored in memory and configured to be executed by the processor to perform a sleep stage determination procedure with the following steps (a) Capturing low-stress physiological signals from a target user; (b) Extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; (c) Analyzing heart rate variability based on beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; (d) To implement the classification of physiological characteristics using a pre-trained machine learning model to determine the sleep stage determination results of the target user. [3] Computer-readable storage medium in which a computer program is stored, characterized by that the steps of a procedure for determining sleep stages include the following steps (a) Capturing low-stress physiological signals from a target user; (b) Extracting heartbeat features from the physiological signals based on the beat-to-beat interval to obtain a beat-to-beat interval sequence; where the beat-to-beat interval refers to the time interval between two successive heartbeats; (c) Analyzing heart rate variability based on beat-to-beat interval sequence and extracting physiological characteristics of the target user from multiple perspectives; (d) Classifying physiological features using a pre-trained machine learning model to determine the sleep stage determination results of the target user. This will be realized when the computer program is executed by the processor.

Images