TW202400076A - Evaluation method of sleep quality and computing apparatus related to sleep quality - Google Patents

Abstract

An evaluation method of sleep quality and a computing apparatus related to sleep quality are provided. In the evaluation method, sensing data is obtained. The sensing data is generated based on radar echo. The sensing data is transformed into feature data. The feature data includes statistics of multiple feature points on the waveform of the radar echo. The sleep quality information is determined according to the feature data. The sleep quality information is related to whether the sleep quality is good or bad. Accordingly, sleep quality could be evaluated through non-touch sensing.

Description

Sleep quality assessment method and sleep quality-related computing device

The present invention relates to a data analysis technology, and in particular to a sleep quality evaluation method and a computing device related to sleep quality.

Sleep apnea refers to the symptom of involuntary weakening or even stopping of breathing during sleep. The cessation of breathing is usually unnoticeable until the body is severely deprived of oxygen and may wake up from discomfort. However, hypoxia may harm the body, and patients may even die suddenly from cardiovascular disease. People with sleep apnea are often unaware of the symptoms. Symptoms can only be discovered unless the patient goes to the hospital for detection and diagnosis using special equipment.

In view of this, embodiments of the present invention provide a method for evaluating sleep quality and a computing device related to sleep quality, which can easily detect sleep quality.

The sleep quality evaluation method according to the embodiment of the present invention includes (but is not limited to) the following steps: obtaining sensing data. This sensing data is generated based on radar echoes. Convert sensing data into feature data. This feature data includes the statistics of multiple feature points on the waveform of the radar echo. Sleep quality information is determined based on characteristic data. This sleep quality information is related to the degree of sleep quality.

Computing devices related to sleep quality in embodiments of the present invention include (but are not limited to) memories and processors. Memory is used to store program code. The processor is coupled to the memory. The processor loads the program code to execute: obtain sensing data, convert the sensing data into characteristic data, and determine sleep quality information based on the characteristic data. This sensing data is generated based on radar echoes. This feature data includes the statistics of multiple feature points on the waveform of the radar echo. This sleep quality information is related to the degree of sleep quality.

Based on the above, according to the sleep quality evaluation method and sleep quality-related computing device according to the embodiment of the present invention, radar-based sensing data is used to predict sleep quality information. The characteristic data obtained from the self-sensing data corresponds to multiple physiological sleep examinations (Polysomnography, PSG). Multiple physiological sleep examinations can be reflected in respiratory events, and respiratory events are related to the quality of sleep. In this way, sleep quality can be assessed through non-contact sensing.

In order to make the above-mentioned features and advantages of the present invention more obvious and easy to understand, embodiments are given below and described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram of components of the computing device 10 and the radar 50 according to an embodiment of the present invention. Referring to FIG. 1 , the computing device 10 includes (but is not limited to) a memory 11 and a processor 12 . The computing device 10 may be a mobile phone, a tablet computer, a notebook computer, a desktop computer, a voice assistant device, a smart home appliance, a wearable device, a vehicle-mounted device or other electronic devices.

The memory 11 can be any type of fixed or removable random access memory (Radom Access Memory, RAM), read only memory (Read Only Memory, ROM), flash memory (flash memory), traditional hard disk (Hard Disk Drive, HDD), solid-state drive (Solid-State Drive, SSD) or similar components. In one embodiment, the memory 11 is used to store program codes, software modules, configurations, data or files (eg, data, events, information, models, or features), which will be described in detail in subsequent embodiments.

The processor 12 is coupled to the memory 11 . The processor 12 may be a central processing unit (CPU), a graphics processing unit (GPU), or other programmable general-purpose or special-purpose microprocessor (Microprocessor), digital signal processing Digital Signal Processor (DSP), programmable controller, Field Programmable Gate Array (FPGA), Application-Specific Integrated Circuit (ASIC), neural network accelerator or other similar elements or combinations of the above elements. In one embodiment, the processor 12 is used to execute all or part of the operations of the computing device 10 , and can load and execute each program code, software module, file and data stored in the memory 11 . In some embodiments, part of the operations in the method of the embodiment of the present invention may be implemented by different or the same processor 12 .

In one embodiment, processor 12 is connected to radar 50 . For example, the radar 50 is connected to the processor 12 through USB, Thunderbolt, Wi-Fi, Bluetooth or other wired or wireless communication technologies. For another example, the computing device 10 has a built-in radar 50, and the processor 12 and the radar 50 are connected through internal circuits. The radar 50 may be a Frequency Modulated Continuous Wave (FMCW) radar or an Impulse Radio (IR)-Ultra-Wideband (UWB) radar. In one embodiment, radar 50 is used to generate sensing data. Sensing data is generated based on radar echoes. Radar echo refers to an echo signal in which the transmission signal of the radar 50 is reflected by an object (for example, a human body or clothing). The sensing data is the sensing result of the radar 50 . For example, in-phase and/or quadrature signals.

In one embodiment, the frequency of the transmitted signal of the radar 50 may be 24 GHz or other frequencies that may reflect the human body (eg, chest or abdomen).

In an application scenario, the radar 50 can be placed at the head of the bed, beside the bed, or at the foot of the bed, and transmits signals toward the chest or abdomen of the human body to detect the ups and downs of the chest or abdomen. However, the installation position and orientation of the radar 50 can still be changed according to actual needs, and are not limited by the embodiment of the present invention.

In the following, the method described in the embodiment of the present invention will be described with reference to various devices, components and modules in the computing device 10 and the radar 50 . Each process of this method can be adjusted according to the implementation situation, and is not limited to this.

FIG. 2 is a flow chart of a sleep quality evaluation method according to an embodiment of the present invention. Referring to FIG. 2 , the processor 12 obtains sensing data (step S210 ). Specifically, the sensing data is generated based on radar echoes. For example, the radar 50 emits a continuous wave signal, and the continuous wave signal is reflected by the chest or abdomen to form a radar echo. The radar 50 can receive the radar echo and generate sensing data accordingly. The sensing data takes in-phase (I) and quadrature (Q) signals as an example. The in-phase signal I1 is [0.164144 0.179153 0.194716 ... 1.600188 1.590467 1.586891], and the quadrature signal Q1 is [2.295545 2.278471 2.270613 ... 1.0 31502 1.027573 1.049331 ].

In one embodiment, the processor 12 may accumulate sensing data for a period of time. This period of time is, for example, 1, 5 or 8 hours.

The processor 12 converts the sensing data into characteristic data (step S220). In one embodiment, the characteristic data includes variation between two channels or within a single channel in the sensing data. These two channels can be in-phase and quadrature signals. The mathematical expression of the variation is: …(1) Cov is the variation number, X and Y are either in-phase or quadrature signals, is the average value of X, is the average value of Y. Taking the aforementioned in-phase signal I1 and quadrature signal Q1 as an example, the variation is -0.21645484961728612.

In one embodiment, the characteristic data includes entropy of the sensed data. In information theory, entropy refers to the average amount of information contained in each message received. It is a measure of uncertainty, and entropy is larger as the source of the message is more random. Features based on entropy include relative entropy, conditional entropy, mutual information, information entropy, Shannon entropy, and block entropy. entropy.

Taking Shannon entropy as an example, the entropy H of a random variable X (with the value range of x={x ₁ , …, x _n }) is defined as: ...(2), P _X is the probability mass function of the random variable X, and b is the base used for logarithms.

Taking the aforementioned in-phase signal I1 and quadrature signal Q1 as an example, their conditional entropy is 1.4112874013149717.

In one embodiment, the feature data includes statistics of multiple feature points on the waveform of the radar echo. Feature points may be peak values and/or valley values in the waveform. For example, FIG. 3 is a schematic diagram of waveform-related characteristic data according to an embodiment of the present invention. Please refer to Figure 3. This waveform includes peak values P1, P2 and valley value V1. The peak values P1 and P2 can be the maximum values in one or more periods. The valley value V1 can be the minimum value within one or more periods.

Furthermore, the statistic may be an interval between two feature points, a variation of the interval, and/or the total number of those feature points. Taking Figure 3 as an example, the statistic is the interval I _PP between the two peak values P1. However, the statistic may also be the interval between the valley value V1 and another adjacent valley value (not shown), the interval between the peak value P1/P2 and the valley value V1, or the interval between two specified points in the waveform. The variation of intervals is, for example, the difference between two or more intervals. For example, the difference between the interval I _PP and another interval I _PP (not shown in the figure, for example, the interval between the peak P2 and the next peak). The total number of feature points is, for example, the total number of peaks and/or valleys within a period of time (eg, 1000 sampling points or 3 hours).

In one embodiment, the processor 12 can determine the statistics of the waveforms of the in-phase and quadrature signals respectively, and can also take the average of the statistics of the two signals as the characteristic data.

In one embodiment, the characteristic data includes a trend of the waveform, and the trend is the intensity change of the waveform without regular characteristics. Regular characteristics can be periodic changes in the waveform. For example, a sine wave signal increases repeatedly from zero to a maximum value, drops from a maximum value to a minimum value, and then increases from a minimum value to zero. After removing the regular characteristics, the trend of the waveform is left. that is, intensity changes. The processor 12 abstracts the trend (that is, eliminates the interference of the absolute intensity of the signal), which can be used as characteristic data describing sleep quality (eg, respiratory events, or sleep events).

For example, FIG. 4 is a schematic diagram of trend-related feature data according to an embodiment of the present invention. Please refer to Figure 4. Radar echoes can be divided into trends and patterns. The trend is a linear function, and the slope of the linear function is positive, so the trend of this waveform is that the intensity gradually increases. The pattern is a string wave. In another embodiment, the trend can also be a curve. In addition, taking the programming language and the aforementioned in-phase signal I1 as an example, python's package: seasonal_decompose can divide the maximum value in the trend of the in-phase signal I1 by the minimum value, and obtain: 1.207502488 (as a representative value of the trend).

In one embodiment, the processor 12 can determine the trends of the waveforms of the in-phase and quadrature signals respectively, and can also take the average of the trends of the two signals as the characteristic data.

In one embodiment, the processor 12 may select one or more of statistics, variation, entropy and trend of the aforementioned sensing data as characteristic data.

Referring to FIG. 2 , the processor 12 determines sleep quality information based on the characteristic data (step S230 ). Specifically, sleep quality information relates to the degree of sleep quality. In one embodiment, the sleep quality information includes respiratory events. For example, normal breathing, hypopnea/hypopnea (hypopnea), shallow slow breathing (Flow Limitation), obstructed breathing, awake or apnea (apnea) events. Apnea is defined as the complete cessation of breathing, pause in breathing, and interruption of airflow for at least 10 seconds during sleep, which is called an event (i.e., a sleep apnea). Hypopnea breathing is defined as an abnormal breathing pattern. For adults, it lasts for at least 10 seconds at a time. The air flow and respiratory movement of the chest and abdomen are reduced to only 30% to 50% compared with normal conditions. At the same time, The oxygen saturation in the blood is reduced by at least 4%. Shallow breathing is defined as an abnormal breathing pattern in which the airflow is lower than normal due to partial obstruction of the airway.

The processor 12 can predict respiratory events based on the characteristic data. The characteristic data in the embodiment of the present invention are characteristics that are better able to distinguish respiratory events and are obtained by comparing multiple physiological sleep examinations (Polysomnography, PSG) (for example, respiratory airflow, chest movement, abdominal muscle behavior, or electroencephalogram). However, unlike identifying respiratory events based on PSG, embodiments of the present invention identify respiratory events based on radar characteristic data.

In one embodiment, the processor 12 can predict respiratory events through a machine learning model. Machine learning models are trained to understand correlations between feature data and respiratory events. Machine learning models are, for example, based on Deep Neural Decision Tree (DNDT), Deep Learning Neural Network (Temporal Convolutional Network, TCN), Decision Tree (Decision Tree), Random Forest (Random Forest) or other machine learning algorithms. Law. Deep learning neural networks are, for example, Temporal Convolutional Network (TCN) and Convolutional Neural Network (CNN). DNDT is a strategy that mixes deep learning and decision trees. Machine learning algorithms can analyze training samples to obtain patterns from them, and then use patterns to predict unknown data. For example, the machine learning model establishes characteristic data (i.e., input to the model) and respiratory events (i.e., model's input) based on labeled samples (e.g., characteristic data of known hypoventilation events, or characteristic data of known normal respiratory events). output) between the nodes in the hidden layer. The machine learning model is a model constructed after learning, and can make inferences based on the evaluation data (for example, feature data).

For example, CNN can learn image-related features, and TCN can learn temporal features. DNDT combines the concepts of decision trees and deep learning in the field of machine learning. Traditional decision trees cannot optimize each tree node, resulting in limitations in judgment. However, DNDT allows the learning machine to perform deep learning to determine the weight of each node of the decision tree.

For example, FIG. 5 is a schematic diagram of a deep neural decision tree (DNDT) according to an embodiment of the present invention. Please refer to Figure 5. During the training process of the model, deep learning is used to continuously change the weight of each node in the decision tree (as shown in the figure, the weight between the input to the discarding layer (Binning Layer) or the Kronecker Product (Kron Product) ) to the output) to further reduce the loss value, thereby achieving the purpose of training weights. The dropout layer is used to implement differentiable grouping functions, and the Kronecker product is used to determine which leaf value in the decision tree to make predictions. Therefore, the characteristic data learned through the DNDT architecture can be further used to determine whether it is normal sleep or whether a sleep/breathing event occurs.

Taking the above-mentioned in-phase signal I1 as an example, the processor 12 can directly convert the in-phase signal I1 into a two-dimensional matrix (for example, the matrix size is 40 x 50 or 30 x 50) (as the input of the model) to learn the machine learning model. Alternatively, the form of the characteristic data may vary according to different time sizes (for example, 100, 1500, 2000 or 2500 sampling points but not limited to this). In some application scenarios, the longer the time or the greater the number of sampling points, the higher the accuracy, but it is not limited to this. Alternatively, the processor 12 may use table-type feature data (as input to the model) to learn the machine learning model. For example, convert the characteristic data of these time series into numerical values and organize them into tables as follows: Table (1) Variation number of in-phase channel Variation number of orthogonal channel Variation between in-phase and orthogonal channels entropy Peak to peak interval Trend 0.058919 0.128993 0.033093 0.58167 0.052821 1.052752 0.049791 0.099063 0.02131 0.478866 0.011539 1.043053 0.013008 0.097488 -0.00206 0.622213 0.006285 1.01288 0.010046 0.114137 -0.008 0.669016 0.000769 1.007484 0.046572 0.138222 0.018872 0.648689 0.024544 1.044307

Figure 6 is a schematic diagram of sensing data and events according to an embodiment of the present invention. Please refer to Figure 6. Event E1 is a respiratory hypoventilation event, and the sensing data and PSG can reflect obvious rapid fluctuations. For example, the end of event E1 has a higher or lower value. Event E2 is a normal sleep event, so the waveforms of the sensing data and PSG change roughly regularly.

The processor 12 can count the number and/or duration of specific respiratory events within a period of time (eg, 2, 5, or 8 hours) as sleep quality information. If the statistic of normal sleep events is higher, the sleep quality is better (for example, the degree of excellence is higher, and high represents superior); if the statistic of normal sleep events such as hypopnea and/or apnea is higher, the sleep quality is poor ( The lower the degree of excellence, and lower represents inferiority).

In one embodiment, the sleep quality information includes sleep statistical indicators. The sleep statistical index is the Respiratory Disturbance Index (RDI) or the Apnea-Hypopnea Index (AHI). RDI is the number of interrupted breathing during sleep, and some people use AHI directly. Under the same measurement conditions, the RDI index will be slightly larger than the AHI index. According to the standards of the American Sleep Association, an AHI below 5 is normal, an AHI of 5 to 14 is mild, an AHI of 15 to 29 is moderate, and an AHI of 30 or above is severe sleep apnea. That is to say, the lower the sleep statistical index, the higher the degree of sleep quality, and high means excellent; the higher the sleep statistical index, the lower the degree of sleep quality, and low means poor.

In one embodiment, the processor 12 may determine sleep statistical indicators based on predicted respiratory events. The processor 12 may aggregate predictions of previous respiratory events (eg, the output of a machine learning model) over a period of time (eg, 3 hours, 5 hours, or 8 hours) and generate predicted sleep statistics. For example, the number of specific respiratory events divided by the statistical time.

For example, Table (2) is the correspondence between time points (for example, every minute, every 30 minutes, or every hour) and prediction results: Table (2) time point Prediction of respiratory events 1 0 2 1 3 1 4 0 … 0 … … N (positive integer) 0 Where "0" means no event, and "1" means there is an event. The AHI can be obtained by dividing the number of respiratory hypopnea events by the statistical time. That is, the frequency of 1 occurring per unit time. In addition, the prediction results of each "1" can be compared and verified with PSG to improve accuracy.

In order to verify whether the RDI-like value (i.e., sleep statistical index) produced by the embodiment of the present invention can be close to the real RDI value, the data of 103 people who underwent sleep testing in the sleep center in the clinical research case were actually collected, and these data were used to conduct Verify. Figure 7 is a schematic diagram of indicator verification according to an embodiment of the present invention. Please refer to Figure 7. The trend and accuracy analysis of the similar RDI (expressed as bRDI on the Y-axis) obtained by the embodiment of the present invention and the RDI judged by the sleep technician (expressed as the RDI on the X-axis) are performed (Table (3)). It can be seen from Figure 7 that the RDI-like value obtained by the embodiment of the present invention is positively correlated with the real RDI value, and the correlation degree (for example, 0.7481) is greater than 0.7.

In addition, clinically, RDI greater than or equal to 15/hour (h) and greater than or equal to 30/h will be defined as having moderate or severe respiratory arrest symptoms, and the comparison results can be obtained in Table (3): table 3) hit rate Set RDI greater than or equal to 15/h as positive (positive) Set RDI greater than or equal to 30/h as positive True Positive 49/64 (76.56%) 36/40 (90.00%) True Negative 32/39 (82.05%) 51/63 (80.95%) Correlation between bRDI and actual RDI 0.7481 0.7481 True positives are the proportions that are judged to be positive by the embodiment of the present invention and are actually positive, and true negatives are the proportions that are judged to be negative by the embodiment of the present invention and are actually negative. It can be seen that the proportion of correct judgments as positive (for example, RDI is greater than 15 times per hour or 30 times per hour) exceeds 75%, and the proportion of correct judgments as negative (for example, RDI is less than 15 times per hour or 30 times per hour) The proportion exceeds 80%.

In one embodiment, the processor 12 can predict sleep statistical indicators based on the characteristic data. For example, the processor 12 additionally trains another machine learning model and uses it to understand the correlation between the feature data and the predicted sleep statistics. The introduction of the machine learning model can be referred to the previous description and will not be repeated here. For example, a machine learning model establishes a relationship between feature data (i.e., input to the model) and sleep statistical indicators (i.e., output of the model) based on labeled samples (e.g., characteristic data of known RDI, or characteristic data of known AHI). The association between nodes in the hidden layer. Since the characteristic data in the embodiment of the present invention can be used to distinguish respiratory events and the sleep statistical indicators are obtained based on respiratory events (for example, the number of specific one or more respiratory events divided by the statistical time), it can be proven that the characteristic data can be used to Predict sleep statistics.

To sum up, in the sleep quality evaluation method and sleep quality-related computing device according to the embodiment of the present invention, the characteristic data (for example, related to variation, entropy, waveform and/or related to the radar sensing data) is converted. or trend) to determine the quality of sleep. In this way, sleep quality can be assessed using non-contact sensing.

Although the present invention has been disclosed above through embodiments, they are not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some modifications and modifications without departing from the spirit and scope of the present invention. Therefore, The protection scope of the present invention shall be determined by the appended patent application scope.

10: Motion device 11: Memory 12: Processor 50: Radar S210~S230: Steps P1, P2: Peak value V1: Valley value I _PP : Interval E1, E2: Event

FIG. 1 is a component block diagram of a computing device and a radar according to an embodiment of the present invention. FIG. 2 is a flow chart of a sleep quality evaluation method according to an embodiment of the present invention. FIG. 3 is a schematic diagram of waveform-related characteristic data according to an embodiment of the present invention. FIG. 4 is a schematic diagram of trend-related feature data according to an embodiment of the present invention. Figure 5 is a schematic diagram of a deep neural decision tree (Deep Neural Decision Tree, DNDT) according to an embodiment of the present invention. Figure 6 is a schematic diagram of sensing data and events according to an embodiment of the present invention. Figure 7 is a schematic diagram of indicator verification according to an embodiment of the present invention.

S210~S230: steps

Claims (20)

An assessment method for sleep quality, including: Obtaining sensing data, wherein the sensing data is generated based on a radar echo; Convert the sensing data into feature data, where the feature data includes a statistic of a plurality of feature points on a waveform of the radar echo; and A piece of sleep quality information is determined based on the characteristic data, wherein the sleep quality information is related to a degree of sleep quality. The sleep quality evaluation method as described in claim 1, wherein the feature points include at least one of a peak value and a valley value, and the statistic includes a value between the two feature points. At least one of an interval, a variation of the interval, and a total number of the feature points. The method for evaluating sleep quality as described in claim 1, wherein the characteristic data further includes a variation between two channels or within a single channel in the sensing data. The sleep quality evaluation method as described in claim 1, wherein the characteristic data further includes an entropy of the sensing data. The sleep quality evaluation method as described in claim 1, wherein the characteristic data further includes a trend of the waveform, and the trend is an intensity change of the waveform without regular characteristics. The sleep quality assessment method as described in claim 1, wherein the sleep quality information includes a respiratory event, and the step of determining the sleep quality information based on the characteristic data includes: The respiratory event is predicted based on the characteristic data. The sleep quality assessment method as described in claim 6, wherein the step of predicting the respiratory event based on the characteristic data includes: The respiratory event is predicted through a machine learning model, wherein the machine learning model is trained to understand the correlation between the feature data and the respiratory event. The sleep quality evaluation method as described in claim 7, wherein the machine learning model is based on a deep neural decision tree (Deep Neural Decision Tree, DNDT), a deep learning neural network (Temporal Convolutional Network, TCN) and a decision-making One of the Decision Trees, and the breathing event is a normal breathing, a hypopnea (hypopnea), a shallow slow breathing (Flow Limitation), an obstructed breathing, an awake or an apnea event. The method for evaluating sleep quality as described in claim 6, wherein the sleep quality information further includes a sleep statistical index, and the sleep statistical index is a sleep disordered breathing index (Respiratory Disturbance Index, RDI) or an apnea-hypopnea index ( Apnea-Hypopnea Index (AHI), and the steps to predict the respiratory event based on this characteristic data include: The sleep statistical index is determined based on the predicted respiratory event. The method for evaluating sleep quality as described in claim 1, wherein the sleep quality information includes a sleep statistical index, the sleep statistical index is a sleep disordered breathing index or an apnea-hypopnea index, and the sleep quality information is determined based on the characteristic data The steps for quality information include: The sleep statistical index is predicted based on the characteristic data. A computing device related to sleep quality, including: a memory to store a program code; and A processor, coupled to the memory, loads the program code to execute: Obtaining sensing data, wherein the sensing data is generated based on a radar echo; Convert the sensing data into feature data, wherein the feature data includes a statistic of a plurality of feature points on a waveform of the radar echo; and A piece of sleep quality information is determined based on the characteristic data, wherein the sleep quality information is related to a degree of sleep quality. The computing device related to sleep quality as described in claim 11, wherein the feature points include at least one of a peak value and a valley value, and the statistics include an interval between two of the feature points, the interval At least one of the change amount and the total number of the feature points. The computing device related to sleep quality as described in claim 11, wherein the characteristic data further includes a variation between two channels or within a single channel in the sensing data. As claimed in claim 11, the computing device related to sleep quality, wherein the characteristic data further includes an entropy of the sensing data. As claimed in claim 11, the computing device related to sleep quality, wherein the characteristic data further includes a trend of the waveform, and the trend is an intensity change of the waveform without regular characteristics. The computing device related to sleep quality as described in claim 11, wherein the sleep quality information includes a breathing event, and the processor further executes: The respiratory event is predicted based on the characteristic data. The computing device related to sleep quality as described in claim 16, wherein the processor further executes: The respiratory event is predicted through a machine learning model, wherein the machine learning model is trained to understand the correlation between the feature data and the respiratory event. The computing device related to sleep quality as described in claim 17, wherein the machine learning model is based on one of a deep neural decision tree, a deep learning neural network and a decision tree, and the breathing event is a normal breathing, a hypopnea, a shallow slow breathing, an obstructed breathing, an awake or an no breathing event. The computing device related to sleep quality as described in claim 16, wherein the sleep quality information further includes a sleep statistical index, the sleep statistical index is a sleep apnea index or an apnea-hypopnea index, and the processor further implement: The sleep statistical index is determined based on the predicted respiratory event. The computing device related to sleep quality as described in claim 11, wherein the sleep quality information includes a sleep statistical index, the sleep statistical index is a sleep apnea index or an apnea hypopnea index, and the processor further executes : The sleep statistical index is predicted based on the characteristic data.

Images