CN115137308A - Method for improving accuracy of in-out sleep detection in sleep algorithm of intelligent wearable device - Google Patents

Abstract

The invention discloses a method for improving the accuracy of sleep entrance and exit detection in a sleep algorithm of intelligent wearable equipment, which comprises the following steps: the method comprises the steps of obtaining corresponding initial acceleration signals and initial heart rate signals through an acceleration sensor and a heart rate sensor, obtaining acceleration signals S1 and heart rate signals S2 after signal processing, obtaining corresponding signal characteristics through analysis processing of the acceleration signals S1 and the heart rate signals S2, preliminarily judging whether the state of a user is in a sleeping state or out of the sleeping state according to a sleep algorithm of the intelligent device, introducing parameters of the using state of a mobile phone when the in-sleeping state and out-sleeping state of the user are detected in the sleep algorithm, improving the accuracy of the intelligent device on in-sleeping and-out time detection, adjusting the in-sleeping and-out time in the sleep algorithm according to feedback of the state of the mobile phone, updating model parameters in the sleep algorithm according to the signal characteristics of the acceleration signals S1 and the heart rate signals S2 in the current state, and finally achieving the purpose of higher accuracy of in-sleeping and-out detection.

Description

Translated fromChinese

一种提高智能穿戴设备睡眠算法中出入睡检测准确性的方法A method for improving the accuracy of falling asleep detection in sleep algorithms of smart wearable devices

技术领域：Technical field:

本发明涉及智能穿戴设备技术领域，特指一种提高智能穿戴设备睡眠算法中出入睡检测准确性的方法。The invention relates to the technical field of smart wearable devices, in particular to a method for improving the accuracy of falling asleep detection in a sleep algorithm of smart wearable devices.

背景技术：Background technique:

随着智能穿戴设备的发展，智能手表、智能手环得到快速的发展。这些智能穿戴设备集成了很多的传感器，可以用来记录用户日常生活中的锻炼、睡眠等相关数据，并且还可以检测用户的身体机能信息。With the development of smart wearable devices, smart watches and smart bracelets have developed rapidly. These smart wearable devices integrate many sensors, which can be used to record the user's daily exercise, sleep and other related data, and can also detect the user's physical function information.

目前用于智能穿戴设备中用于检测用户睡眠信息的方式主要是通过检测用户的心率等身体指标参数，以及用户的身体活动参数，判断用户的出入睡状态。见专利申请号为：201810721118.1中国发明专利申请，其公开了“一种用于检测用户睡眠状态的检测系统及检测方法”，其采用的技术方案为：一种用于检测用户睡眠状态的检测系统及检测方法，包括可穿戴智能监测设备，用于监测用户的心率、血氧和体动数据；服务器，所述服务器与可穿戴智能监测设备通信连接，用于接收并储存可穿戴智能监测设备监测到的心率、血氧和体动数据，服务器包括分析模块，所述分析模块用于实现用户的睡眠状态分析。该发明专利申请利用可穿戴智能监测设备在睡眠过程中实时采集心率、血氧和体动参数，采集数据实时传输到服务器端进行分析判断睡眠状态，其分析判断方法是基于体动数据基础上结合心率和血氧数据进行分析，判断用户真实睡眠状态。The method currently used for detecting user sleep information in smart wearable devices is mainly to determine the user's sleep state by detecting the user's heart rate and other body index parameters, as well as the user's physical activity parameters. See the patent application number: 201810721118.1 Chinese invention patent application, which discloses "a detection system and detection method for detecting the sleep state of a user", and the technical solution adopted is: a detection system for detecting the sleep state of a user and a detection method, including a wearable intelligent monitoring device for monitoring the user's heart rate, blood oxygen and body movement data; a server, which is connected in communication with the wearable intelligent monitoring device and used for receiving and storing the wearable intelligent monitoring device monitoring The received heart rate, blood oxygen and body movement data, the server includes an analysis module, and the analysis module is used to analyze the sleep state of the user. The invention patent application uses wearable intelligent monitoring equipment to collect heart rate, blood oxygen and body motion parameters in real time during sleep, and the collected data is transmitted to the server in real time for analysis and judgment of sleep state. The analysis and judgment method is based on the combination of body motion data. Heart rate and blood oxygen data are analyzed to determine the user's real sleep state.

通过上述发明专利申请所述，目前对于智能穿戴设备而言，其睡眠算法中，用于判断用户出入睡状态的方法，就是采集多种不同的参数(如上述发明专利，其为了提高睡眠检测的准确性应用了心率、血氧和体动数据作为判断依据)，根据用户在睡眠状态下参数的变化，确定相应的判断阈值，当参数处于判断阈值中，则智能穿戴设备就会判断用户为睡眠状态，反之智能穿戴设备就会判断用户为非睡眠状态。As mentioned in the above patent application for invention, currently, for smart wearable devices, the method for judging the user's sleep state in the sleep algorithm is to collect a variety of different parameters (such as the above invention patent, which is used to improve sleep detection performance. The accuracy uses heart rate, blood oxygen and body motion data as the judgment basis), and determines the corresponding judgment threshold according to the change of the user's parameters in the sleep state. When the parameter is within the judgment threshold, the smart wearable device will judge that the user is sleeping. On the contrary, the smart wearable device will judge that the user is in a non-sleep state.

通过心率、血氧和体动数据等多种参数作为判断依据，虽然有助于提高判断睡眠状态的准确性，但是仍存在一定不足。主要体现在以下方面：Using various parameters such as heart rate, blood oxygen and body movement data as the judgment basis, although it helps to improve the accuracy of judging the sleep state, there are still some shortcomings. Mainly reflected in the following aspects:

大多数人在睡觉前通常会在床上长时间的进行较小的活动，例如躺着看书、看手机等；在醒来后也会在床上进行长时间的进行较小的活动，例如躺着看书、看手机等。此时，人体的活动动作非常小，心率等参数变化并不显著，这就容易导致智能手环、手表等穿戴设备在检测用户睡眠状态时出错，例如，通过其现有的算法，错误的认为用户已经入睡(实质仍在床上看手机)，或者通过其现有的算法，错误的认为用户仍然在睡眠中(实质已经醒来看手机)。Most people usually do small activities in bed for a long time before going to bed, such as lying down to read a book, looking at a mobile phone, etc.; after waking up, they also do small activities in bed for a long time, such as lying down and reading a book , look at the phone, etc. At this time, the movement of the human body is very small, and the changes in parameters such as heart rate are not significant, which easily leads to errors in the detection of the user's sleep state by wearable devices such as smart bracelets and watches. The user has fallen asleep (essentially still looking at the phone in bed), or by its existing algorithm, it incorrectly believes that the user is still sleeping (essentially woke up to look at the phone).

另一方面，随着智能手机的发展，各种智能设备在使用时均和手机实现了互联，通过APP可以实现智能设备与手机的互联。On the other hand, with the development of smart phones, all kinds of smart devices are interconnected with mobile phones when in use, and the interconnection between smart devices and mobile phones can be realized through APP.

综上所述，本发明人针对目前智能手表、手环等智能穿戴设备睡眠算法中存在的不足，提出以下技术方案。To sum up, the inventors of the present invention propose the following technical solutions in view of the deficiencies in the sleep algorithms of current smart wearable devices such as smart watches and wristbands.

发明内容：Invention content:

本发明所要解决的技术问题就在于克服现有技术的不足，提供一种提高智能穿戴设备睡眠算法中出入睡检测准确性的方法。The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art, and to provide a method for improving the accuracy of falling asleep detection in a sleep algorithm of an intelligent wearable device.

为了解决上述技术问题，本发明采用了下述技术方案：一种提高智能穿戴设备睡眠算法中出入睡检测准确性的方法，该方法为：通过集成在智能穿戴设备的加速度传感器和心率传感器，获取相应的初始加速度信号和初始心率信号，经过对信号处理后获取加速度信号S1和心率信号S2，通过对加速度信号S1和心率信号S2的分析处理得到对应的信号特征，并根据智能设备睡眠算法，初步判断用户的状态为入睡状态或出睡状态，该方法中引入手机状态作为参数，当初步判断用户为进入入睡状态，用户手机状态为未在玩手机，则判断为：用户进入入睡状态，下一步进入出入睡检测输出状态；当初步判断用户进入入睡状态，而用户玩手机状态为在玩手机，则判断为：用户未进入入睡状态，下一步进入出入睡检测更新状态；当初步判断用户进入出睡状态，而用户玩手机状态为在玩手机，则判断为：用户进入出睡状态，下一步进入出入睡检测输出状态；当初步判断用户未进入出睡状态，而用户玩手机状态为在玩手机，则判断为：用户进入出睡状态，下一步进入出入睡检测输出状态和出入睡检测更新状态；通过上述判断结果，当进入出入睡检测输出状态，直接输出判断的结果；当进入出入睡检测更新状态，将对应状态下加速度信号S1和心率信号S2的信号特征，更新据智能设备睡眠算法的模型参数，经过模型参数不断的更新，不断提高出入睡检测的准确度。In order to solve the above-mentioned technical problems, the present invention adopts the following technical solutions: a method for improving the accuracy of falling asleep detection in the sleep algorithm of smart wearable devices, the method is as follows: through the acceleration sensor and the heart rate sensor integrated in the smart wearable device, obtain Corresponding initial acceleration signal and initial heart rate signal, the acceleration signal S1 and heart rate signal S2 are obtained after signal processing, and the corresponding signal characteristics are obtained by analyzing and processing the acceleration signal S1 and heart rate signal S2, and according to the smart device sleep algorithm, preliminary It is judged that the state of the user is in the state of falling asleep or the state of falling asleep. The method introduces the state of the mobile phone as a parameter. When it is initially determined that the user is in the state of falling asleep and the state of the user's mobile phone is not playing the mobile phone, it is determined that the state of the user is in the state of falling asleep, and the next step Entering the falling asleep detection output state; when it is initially judged that the user has entered the falling asleep state, and the user is playing with the mobile phone, it is judged as: the user has not entered the falling asleep state, and the next step is to enter the falling asleep detection and updating state; Sleeping state, and the user is playing with the mobile phone, it is judged as: the user enters the falling asleep state, and the next step is to enter the falling asleep detection output state; when it is initially judged that the user has not entered the falling asleep state, and the user is playing the mobile phone state is playing The mobile phone is judged as: the user enters the falling asleep state, and the next step is to enter the falling asleep detection output state and the falling asleep detection update state; through the above judgment results, when entering the falling asleep detection output state, the judgment result is directly output; when entering and falling asleep Detect the update state, update the model parameters of the sleep algorithm of the smart device according to the signal characteristics of the acceleration signal S1 and the heart rate signal S2 in the corresponding state, and continuously improve the accuracy of falling asleep detection through continuous updating of the model parameters.

进一步而言，上述技术方案中，所述的加速度信号S1的获取方法为：采集加速度传感器的三轴的加速度数据xACC、yACC、zACC，然后对三轴加速度信号进行矢量合成操作，得到合成加速度信号，对合成加速度信号经过滤波去噪处理，得到加速度信号S1，再计算设定时间窗口内信号的波动信号特征。Further, in the above technical solution, the method for obtaining the acceleration signal S1 is: collecting the three-axis acceleration data xACC, yACC and zACC of the acceleration sensor, and then performing a vector synthesis operation on the three-axis acceleration signal to obtain a synthesized acceleration signal , the synthetic acceleration signal is filtered and denoised to obtain the acceleration signal S1, and then the fluctuation signal characteristics of the signal in the set time window are calculated.

进一步而言，上述技术方案中，所述的加速度信号S1获取方法中，对三轴加速度信号进行矢量合成操作为：将三轴加速度信号的平方和再开方，对应计算公式为：

合成加速度信号经过滤波去噪处理后，找到设定窗口内的最大值和最小值，根据最大值和最小值之差计算合成加速度信号的波动信号特征。Further, in the above technical solution, in the method for obtaining the acceleration signal S1, the vector synthesis operation on the three-axis acceleration signal is as follows: the square sum of the three-axis acceleration signal is re-squared, and the corresponding calculation formula is:

After the synthetic acceleration signal is filtered and denoised, the maximum and minimum values in the set window are found, and the fluctuation signal characteristics of the synthetic acceleration signal are calculated according to the difference between the maximum and minimum values.

进一步而言，上述技术方案中，所述心率信号S2的获取方法为：采集心率传感器的心率波信号，然后对采集到的心率波信号进行滤波去噪处理，得到心率信号S2，再提取心率、心率变异性的波形信号特征。Further, in the above technical solution, the acquisition method of the heart rate signal S2 is: collecting the heart rate wave signal of the heart rate sensor, then filtering and denoising the collected heart rate wave signal to obtain the heart rate signal S2, and then extracting the heart rate, Waveform signal characteristics of heart rate variability.

进一步而言，上述技术方案中，所述的心率、心率变异性的波形信号特征获取方法为：对滤波去噪处理后的心率波信号S2利用谷/峰值检测算法，检测有效的谷/峰值点，根据设定时间窗口内检测到的谷/峰位置，结合采样率,计算得到心率、心率变异性的波形信号特征。Further, in the above technical solution, the method for obtaining the waveform signal features of the heart rate and heart rate variability is as follows: using a valley/peak detection algorithm on the heart rate wave signal S2 after filtering and denoising processing to detect an effective valley/peak point , according to the detected valley/peak position within the set time window, combined with the sampling rate, the waveform signal characteristics of heart rate and heart rate variability are calculated.

进一步而言，上述技术方案中，所述的智能穿戴设备为具有睡眠监测的智能手表、智能手环。Further, in the above technical solution, the smart wearable device is a smart watch or smart bracelet with sleep monitoring.

本发明是利用现有智能穿戴设备与手机之间的互联系统，在睡眠算法中，对用户出入睡状态检测时，引入手机使用状态的参数，从而提高智能设备对出入睡时间检测的准确度，例如，当检测到用户正在玩手机，此时智能设备就会认定用户处于清醒状态，根据手机状态的反馈调整睡眠算法中出入睡时间，同时将当前状态下加速度信号S1和心率信号S2的信号特征，更新睡眠算法中的模型参数，在用户不断使用过程中，算法经过不断的更新学习，令算法中模型参数越来越精确，最终实现对于出入睡检测准确度越来越高的目的。The present invention utilizes the interconnection system between the existing smart wearable device and the mobile phone. In the sleep algorithm, when detecting the user's sleeping state, the parameters of the mobile phone use state are introduced, so as to improve the accuracy of the smart device's detection of the sleeping time. For example, when it is detected that the user is playing with the mobile phone, the smart device will determine that the user is in a awake state, adjust the time to fall asleep in the sleep algorithm according to the feedback of the mobile phone state, and at the same time, the signal characteristics of the acceleration signal S1 and the heart rate signal S2 in the current state are adjusted. , update the model parameters in the sleep algorithm. During the continuous use of the user, the algorithm undergoes continuous update and learning, so that the model parameters in the algorithm become more and more accurate, and finally achieve the purpose of increasing the accuracy of falling asleep detection.

附图说明：Description of drawings:

图1是本发明中睡眠算法的流程图；Fig. 1 is the flow chart of sleep algorithm in the present invention;

图2a-2d是本发明中睡眠算法中出入睡状态判断原理图；2a-2d are schematic diagrams of judging states of falling asleep and falling asleep in the sleep algorithm of the present invention;

图3是本发明一具体实施范式中加速度信号S1的信号图。FIG. 3 is a signal diagram of the acceleration signal S1 in an embodiment of the present invention.

具体实施方式：Detailed ways:

下面结合具体实施例和附图对本发明进一步说明。The present invention will be further described below with reference to specific embodiments and accompanying drawings.

本发明为一种提高智能穿戴设备睡眠算法中出入睡检测准确性的方法，所述的智能穿戴设备可以是智能手表、智能手环等电子设备。这些电子设备中通常提供睡眠监测模块，其睡眠监测模块采用的睡眠算法与目前通常的算法基本相同，通过集成在智能穿戴设备的加速度传感器和心率传感器，获取相应的原始加速度信号和原始心率信号，经过对信号处理后获取加速度信号S1和心率信号S2，通过对加速度信号S1和心率信号S2的分析得到对应的信号特征，并根据设定的模型参数，初步判断用户的状态为入睡状态或出睡状态。The present invention is a method for improving the accuracy of falling asleep detection in a sleep algorithm of a smart wearable device. The smart wearable device can be an electronic device such as a smart watch and a smart bracelet. These electronic devices usually provide a sleep monitoring module. The sleep algorithm used in the sleep monitoring module is basically the same as the current algorithm. Through the acceleration sensor and heart rate sensor integrated in the smart wearable device, the corresponding raw acceleration signal and raw heart rate signal are obtained. After signal processing, the acceleration signal S1 and the heart rate signal S2 are obtained, and the corresponding signal characteristics are obtained through the analysis of the acceleration signal S1 and the heart rate signal S2, and according to the set model parameters, the user's state is preliminarily determined as falling asleep or falling asleep state.

本发明针对该睡眠算法进行了进一步的优化，从而令算法中出入睡检测的准确性可以进一步的提高。简单来说，本发明是在现有的睡眠算法中引入手机状态作为参数，通过对手机状态的识别就可以进一步确认用户是否是入睡或出睡状态。具体而而言，对于用户入睡、出睡的检测包括以下四种情形。The present invention further optimizes the sleep algorithm, so that the accuracy of falling asleep detection in the algorithm can be further improved. To put it simply, the present invention introduces the state of the mobile phone as a parameter in the existing sleep algorithm, and by identifying the state of the mobile phone, it can be further confirmed whether the user is in the state of falling asleep or falling asleep. Specifically, the detection of the user falling asleep and falling asleep includes the following four situations.

1、当初步判断用户为进入入睡状态，用户手机状态为未在玩手机，则判断为：用户进入入睡状态；1. When it is preliminarily judged that the user is in the state of falling asleep and the state of the user's mobile phone is not playing the mobile phone, it is determined that the user is in the state of falling asleep;

2、当初步判断用户进入入睡状态，而用户玩手机状态为在玩手机，则判断为：用户未进入入睡状态；2. When it is preliminarily judged that the user has entered the sleep state, and the user is playing with the mobile phone, it is judged that the user has not entered the sleep state;

3、当初步判断用户进入出睡状态，而用户玩手机状态为在玩手机，则判断为：用户进入出睡状态；3. When it is preliminarily judged that the user has entered the state of falling asleep, and the state of the user playing the mobile phone is playing with the mobile phone, it is judged that the user has entered the state of falling asleep;

4、当初步判断用户未进入出睡状态，而用户玩手机状态为在玩手机，则判断为：用户进入出睡状态；4. When it is preliminarily judged that the user has not entered the sleep state, and the user is playing with the mobile phone, it is judged that the user has entered the sleep state;

对于第1、3种情况而言，最终判断出用户进入入睡和出睡状态，则下一步直接进入出入睡检测输出状态，输出用户的入睡或出睡状态至智能设备处理单元，由智能设备记录用户的睡眠数据。即，通过上述检测判断，当进入出入睡检测输出状态，直接输出判断的结果。For the first and third cases, it is finally determined that the user has entered the state of falling asleep and falling asleep, then the next step is to directly enter the output state of falling asleep and falling asleep, and output the user's falling asleep or falling asleep state to the smart device processing unit, which is recorded by the smart device. User's sleep data. That is, through the above-mentioned detection judgment, when entering the falling asleep detection output state, the judgment result is directly output.

对于第2种的情况而言，虽然根据睡眠算法初步判断用户进入入睡状态，但是通过对用户手机状态的校正，判断出用户实际并未进入入睡状态；对于第4种情况而言，虽然根据睡眠算法初步未判断出用户进入出睡状态，但是通过对用户手机状态的校正，判断出用户实际已经进入出睡状态。For the second case, although it is preliminarily judged that the user has entered the sleep state according to the sleep algorithm, it is judged that the user has not actually entered the sleep state through the correction of the user's mobile phone state; for the fourth case, although according to the sleep state The algorithm does not initially determine that the user has entered the state of falling asleep, but through the correction of the state of the user's mobile phone, it is determined that the user has actually entered the state of falling asleep.

当进入出入睡检测更新状态，即将对应状态下加速度信号S1和心率信号S2的信号特征，更新睡眠算法中设定的模型参数，这样经过模型参数不断的更新，不断提高出入睡检测的准确度。When entering the sleep-fall detection update state, the signal characteristics of the acceleration signal S1 and the heart rate signal S2 in the corresponding state are updated to the model parameters set in the sleep algorithm, so that the accuracy of sleep-fall detection is continuously improved through the continuous updating of the model parameters.

由于本发明引入了手机状态参数，通过手机状态参数可以校正睡眠算法中的误判，并将不断更新算中对应的加速度信号S1和心率信号S2的信号特征，经过不断算法模型的不断更新学习，睡眠算法得到的结果可以越来越精准。最后不再引入手机状态参数后，算法也可以根据不断更新学习，得到更加精确的结果。Because the mobile phone state parameter is introduced in the present invention, the misjudgment in the sleep algorithm can be corrected through the mobile phone state parameter, and the signal characteristics of the corresponding acceleration signal S1 and the heart rate signal S2 in the calculation will be continuously updated. The results obtained by the sleep algorithm can be more and more accurate. Finally, after the mobile phone state parameters are no longer introduced, the algorithm can also learn based on continuous updates to obtain more accurate results.

结合图1所示，本发明的睡眠算法中加速度信号S1的获取方法为：With reference to Figure 1, the method for acquiring the acceleration signal S1 in the sleep algorithm of the present invention is:

首先，采集原始的加速度(gsensor)信号。原始的加速度传感器包括X/Y/Z三个轴的加速度数据xACC、yACC、zACC。First, the raw acceleration (gsensor) signal is acquired. The original acceleration sensor includes acceleration data xACC, yACC, and zACC of the three axes of X/Y/Z.

然后，对三轴加速度信号进行矢量合成操作，得到合成加速度信号。对三轴加速度信号进行矢量合成操作为：将三轴加速度信号的平方和再开方，对应计算公式为：

Then, a vector synthesis operation is performed on the three-axis acceleration signal to obtain a synthesized acceleration signal. The vector synthesis operation for the three-axis acceleration signal is: square the sum of the squares of the three-axis acceleration signal, and the corresponding calculation formula is:

接着，对合成加速度信号进行信号处理，并进行特征提取。该步骤中，对于信号的处理主要为：该经过滤波去噪处理，得到加速度信号S1。通常进行3Hz(可根据实际情况调整)的低通滤波得到处理加速度信号S1。所述的特征提取是计算设定窗长内信号的波动信号特征，并提取该波动信号特征。采用的方法为：合成加速度信号经过滤波去噪处理后，找到设定窗口内的最大值和最小值，根据最大值和最小值之差计算合成加速度信号波动信号特征。这种合成加速度信号波动信号特征是作为睡眠算法中的一个参数，通常当该信号特征位于某一阈值范围内可判断为入睡状态，当该信号特征位于另外一阈值范围内可判断为出睡状态。为了增加准确性，当连续的合成加速度信号波动信号特征均位于设定阈值范围内时，则做出相应的判断。Next, signal processing is performed on the synthesized acceleration signal, and feature extraction is performed. In this step, the signal processing is mainly as follows: the acceleration signal S1 is obtained after filtering and denoising processing. Usually, a low-pass filter of 3 Hz (which can be adjusted according to the actual situation) is performed to obtain the processed acceleration signal S1. The feature extraction is to calculate the fluctuation signal characteristics of the signal within the set window length, and extract the fluctuation signal characteristics. The method adopted is: after the synthetic acceleration signal is filtered and denoised, the maximum and minimum values in the set window are found, and the fluctuation signal characteristics of the synthetic acceleration signal are calculated according to the difference between the maximum and minimum values. This synthetic acceleration signal fluctuation signal feature is used as a parameter in the sleep algorithm. Usually, when the signal feature is within a certain threshold range, it can be judged as falling asleep, and when the signal feature is within another threshold range, it can be judged as falling asleep. . In order to increase the accuracy, when the continuous synthetic acceleration signal fluctuation signal features are all within the set threshold range, a corresponding judgment is made.

结合图1所示，本发明的睡眠算法中所述心率信号S2的获取方法为：With reference to Fig. 1, the method for obtaining the heart rate signal S2 in the sleep algorithm of the present invention is as follows:

首先，采集心率传感器的原始心率波(ppg)信号，现有的心率传感器多采用光电容积脉搏波传感器，其是通过光信号检测人体脉搏信号,从而获取用户心率信号。First, the original heart rate wave (ppg) signal of the heart rate sensor is collected. The existing heart rate sensors mostly use a photoplethysmography sensor, which detects the human pulse signal through the optical signal, thereby obtaining the user's heart rate signal.

然后对采集到的心率波信号进行滤波去噪处理，通常进行3Hz(可根据实际情况调整)的低通滤波，去除杂波，并突出信号中的脉搏波信号部分，得到心率信号S2，Then the collected heart rate wave signal is filtered and denoised, usually a low-pass filter of 3Hz (which can be adjusted according to the actual situation) is performed to remove clutter, and the pulse wave signal part in the signal is highlighted to obtain the heart rate signal S2,

接着，再提取心率、心率变异性的波形信号特征。本步骤中，所述的心率、心率变异性的波形信号特征获取方法为：对滤波去噪处理后的心率波信号S2利用谷(峰)值检测算法，检测有效的谷(峰)值点，根据设定窗宽内检测到的谷(峰)位置，结合采样率,计算心率、心率变异性的波形信号特征。例如，在单位时间T内检测信号的有效波峰位置，根据采样的频率分别于T1、T2……时间点检测到对应有效波峰，则根据检测频率就可以得到用户的心率，以及心率变化的波形信号特征。心率、心率变异性的波形信号特征作为睡眠算法中的另外一个参数，当其处于某一阈值范围内，可判断为入睡状态。当该信号特征位于另外一阈值范围内可判断为出睡状态。为了增加准确性，当连续的信号特征均位于设定阈值范围内时，则做出相应的判断。Next, the waveform signal features of heart rate and heart rate variability are extracted. In this step, the method for obtaining the waveform signal features of the heart rate and heart rate variability is as follows: using a valley (peak) value detection algorithm on the heart rate wave signal S2 after filtering and denoising processing, to detect an effective valley (peak) value point, According to the detected valley (peak) position within the set window width, combined with the sampling rate, the waveform signal characteristics of heart rate and heart rate variability are calculated. For example, the effective peak position of the signal is detected within the unit time T, and the corresponding effective peak is detected at T1, T2... feature. The waveform signal features of heart rate and heart rate variability are another parameter in the sleep algorithm. When it is within a certain threshold range, it can be judged as falling asleep. When the signal characteristic is within another threshold range, it can be determined that the state of falling asleep. In order to increase the accuracy, when the continuous signal features are all within the set threshold range, the corresponding judgment is made.

见图1所示，通过上述算法中得到合成加速度信号波动信号特征和心率、心率变异性的波形信号特征，输入到算法出入睡检测模块中，初步判断用户的睡眠状态，当两种信号特征均做出相同的判断，则得到初步的睡眠状态判断结果。初步判断结果将与手机检测玩手机状态的参数一起输入综合出入睡检测模块中，进行最终的睡眠状态判断。As shown in Figure 1, through the above algorithm, the fluctuation signal characteristics of the synthetic acceleration signal and the waveform signal characteristics of heart rate and heart rate variability are obtained, and input into the sleep detection module of the algorithm to preliminarily determine the user's sleep state. When the two signal characteristics are both If the same judgment is made, a preliminary sleep state judgment result is obtained. The preliminary judgment result will be input into the integrated sleep and fall detection module together with the parameters of the mobile phone detecting and playing the mobile phone state, and the final sleep state judgment will be made.

见图2a-2d所示，结合前面所述，出入睡检测模块对于用户入睡、出睡的检测最终包括以下四种情形。As shown in Figs. 2a-2d, in combination with the foregoing, the detection of the user falling asleep and falling asleep by the falling asleep detection module finally includes the following four situations.

1、算法出入睡检测模块检测为入睡状态，用户手机状态为未在玩手机，则综合出入睡检测模块最终判断为：入睡。1. The algorithm detects that the sleep-in and out-sleep detection module is in the state of falling asleep, and the user's mobile phone state is not playing the mobile phone, then the comprehensive in-sleep and out-sleep detection module finally judges: fall asleep.

2、算法出入睡检测模块检测为入睡状态，用户手机状态为在玩手机，则综合出入睡检测模块最终判断为：未入睡，下一步进入算法出入睡检测参数更新模块，缓存当前数据。2. If the sleep detection module of the algorithm detects that it is in the sleep state, and the user's mobile phone state is playing mobile phone, the comprehensive sleep detection module finally judges that it is not asleep, and the next step is to enter the algorithm sleep detection parameter update module to cache the current data.

3、算法出入睡检测模块检测为出睡状态，用户手机状态为在玩手机，则综合出入睡检测模块最终判断为：出睡。3. The algorithm detects that the falling asleep detection module is in the falling asleep state, and the user's mobile phone state is playing the mobile phone, then the comprehensive falling asleep detection module finally judges that: falling asleep.

4、算法出入睡检测模块未检测为出睡状态，用户手机状态为在玩手机，则综合出入睡检测模块最终判断为：出睡，缓冲当前数据。4. The algorithm does not detect the falling asleep state, and the user's mobile phone state is playing mobile phone, then the comprehensive falling asleep detection module finally judges: falling asleep, buffering the current data.

当最终判断出用户进入入睡和出睡状态，则下一步直接进入睡眠出入睡检测输出状态，输出用户的入睡或出睡状态至智能设备处理单元，由智能设备记录用户的睡眠数据。即，通过上述检测判断，当进入出入睡检测输出状态，直接输出判断的结果。When it is finally determined that the user is in the state of falling asleep and falling asleep, the next step is to directly enter the output state of sleep and falling asleep detection, output the user's falling asleep or falling asleep state to the smart device processing unit, and the smart device records the user's sleep data. That is, through the above-mentioned detection judgment, when entering the falling asleep detection output state, the judgment result is directly output.

当进入算法出入睡检测参数更新状态，即将对应状态下加速度信号S1和心率信号S2的信号特征，更新睡眠算法中设定的模型参数。即，对于缓冲数据提取到合成加速度信号波动特征和心率、心率变异性的波形信号特征，结合缓存的标签数据，将该数据进行在线重新训练模型，更新模型参数。When the algorithm enters the sleep detection parameter update state of the algorithm, the model parameters set in the sleep algorithm are updated according to the signal characteristics of the acceleration signal S1 and the heart rate signal S2 in the corresponding state. That is, for the waveform signal characteristics of synthetic acceleration signal fluctuation characteristics and heart rate and heart rate variability extracted from the buffer data, combined with the cached label data, the data is retrained online, and the model parameters are updated.

见图3所示，这是本发明一具体实施范式中加速度信号S1的信号图。其中，t1为躺在床上开始玩手机的时间；t2为真实入睡时间；t3为睡眠算法出入睡检测模块更新参数后的模型检测到的入睡时间；t4为睡眠算法出入睡检测模块更新参数前检测到的初步入睡时间。As shown in FIG. 3 , which is a signal diagram of the acceleration signal S1 in a specific implementation form of the present invention. Among them, t1 is the time to start playing with the phone while lying on the bed; t2 is the actual time to fall asleep; t3 is the time to fall asleep detected by the model after the parameters have been updated by the sleep algorithm sleep detection module; t4 is the detection before the sleep algorithm sleep detection module updates the parameters to the initial sleep time.

由此可见，采用一般的睡眠算法，其初步判断的入睡时间t4，其距离真实的入睡时间t2提前了较长的时间，经过不断更新参数后的睡眠算法后，最终判断检测的入睡时间为t3,其距离真实的入睡时间t2相对于t4准确性大大提高。It can be seen that, using the general sleep algorithm, the preliminarily determined sleep-on time t4 is a long time ahead of the real sleep-on time t2, and after the sleep algorithm that continuously updates the parameters, the detected sleep-on time is finally determined to be t3 , its distance from the real sleep time t2 is greatly improved compared to t4.

当然，以上所述仅为本发明的具体实施例而已，并非来限制本发明实施范围，凡依本发明申请专利范围所述构造、特征及原理所做的等效变化或修饰，均应包括于本发明申请专利范围内。Of course, the above descriptions are only specific embodiments of the present invention, and are not intended to limit the scope of implementation of the present invention. Any equivalent changes or modifications made in accordance with the structures, features and principles described in the scope of the patent application of the present invention shall be included in the within the scope of the patent application of the present invention.

Claims (7)

1. A method for improving the accuracy of in-and-out sleep detection in a sleep algorithm of intelligent wearable equipment comprises the following steps: through integrated acceleration sensor and the rhythm of the heart sensor at intelligent wearing equipment, acquire corresponding original acceleration signal and original rhythm of the heart signal, through acquire acceleration signal S1 and rhythm of the heart signal S2 after to signal processing, obtain corresponding signal characteristic through the analytic processing to acceleration signal S1 and rhythm of the heart signal S2 to according to smart machine sleep algorithm, tentatively judge user' S state for falling asleep state or the state of falling asleep, its characterized in that:

the method introduces the state of the mobile phone as a parameter,

when the user is judged to be in the sleep state in the initial step and the mobile phone state of the user is not playing the mobile phone, the judgment is that: the user enters a sleeping state, and then enters a sleeping in and out detection output state;

when the user is judged to enter the sleep state at first and the mobile phone playing state of the user is the mobile phone playing state, the judgment is that: if the user does not enter the sleep state, the user enters the sleep in and out detection updating state;

when the user is judged to enter the sleep state in the initial step and the mobile phone playing state of the user is the mobile phone playing state, the judgment is that: the user enters a sleep-out state, and then enters a sleep-in and sleep-out detection output state;

when the user is judged not to fall asleep at the beginning, and the user plays the mobile phone, the judgment is that: the user enters a sleeping-out state, and then enters a sleeping-in/out detection output state and a sleeping-in/out detection updating state;

outputting a judgment result when the user enters a sleep in and out detection output state according to the judgment result; when the user enters the in-and-out sleep detection updating state, the signal characteristics of the acceleration signal S1 and the heart rate signal S2 in the corresponding states are updated according to the model parameters of the sleep algorithm of the intelligent device, and the accuracy of in-and-out sleep detection is continuously improved through continuous updating of the model parameters.

2. The method for improving the accuracy of the out-of-sleep detection in the sleep algorithm of the intelligent wearable device as claimed in claim 1, wherein the method comprises the following steps: the method for acquiring the acceleration signal S1 comprises the following steps: acquiring triaxial acceleration data xACC, yACC and zACC of an acceleration sensor, then carrying out vector synthesis operation on triaxial acceleration signals to obtain synthesized acceleration signals, carrying out filtering and denoising processing on the synthesized acceleration signals to obtain acceleration signals S1, and then calculating the fluctuation signal characteristics of the signals in a set time window.

3. The method for improving the accuracy of the out-of-sleep detection in the sleep algorithm of the intelligent wearable device according to claim 2, wherein the method comprises the following steps: in the method for acquiring the acceleration signal S1, the vector synthesis operation performed on the three-axis acceleration signal is: and (3) re-squaring the sum of squares of the three-axis acceleration signals, wherein the corresponding calculation formula is as follows:

4. the method for improving the accuracy of the out-of-sleep detection in the sleep algorithm of the intelligent wearable device according to claim 2, wherein the method comprises the following steps: and after filtering and denoising the synthesized acceleration signal, finding out the maximum value and the minimum value in a set window, and calculating the fluctuation signal characteristic of the synthesized acceleration signal according to the difference between the maximum value and the minimum value.

5. The method for improving the accuracy of the detection of going out of sleep in the sleep algorithm of the intelligent wearable device as claimed in claim 1, wherein: the method for acquiring the heart rate signal S2 comprises the following steps: the method comprises the steps of collecting heart rate wave signals of a heart rate sensor, then carrying out filtering and denoising processing on the collected heart rate wave signals to obtain heart rate signals S2, and then extracting waveform signal characteristics of heart rate and heart rate variability.

6. The method for improving the accuracy of the out-of-sleep detection in the sleep algorithm of the intelligent wearable device according to claim 5, characterized in that: the method for acquiring the waveform signal characteristics of the heart rate and the heart rate variability comprises the following steps: and detecting effective valley/peak points of the heart rate wave signal S2 subjected to filtering and denoising by using a valley/peak detection algorithm, and calculating to obtain the waveform signal characteristics of the heart rate and the heart rate variability according to the detected valley/peak positions in a set time window and in combination with the sampling rate.

7. The method for improving the accuracy of in-and-out sleep detection in the sleep algorithm of the intelligent wearable device according to any one of claims 1 to 6, wherein the method comprises the following steps: the intelligent wearable device comprises an intelligent watch and an intelligent bracelet, wherein the intelligent watch and the intelligent bracelet are provided with sleep monitoring functions.

Images