除副交感神经活跃这一表现外，人在睡眠过程中体动状况相对于觉醒期体动状况

Except for the performance of active parasympathetic nervous system, people's body movement during sleep is much less than that of wakefulness, especially the periodic leg movement signals and turnover signals are more significant [8]. Obvious differences can be seen in the characteristics of periodic leg movements in different sleep stages. The relative frequency, duration, and wake-causing effects all decrease with the gradual deepening of sleep, while the interval between leg movements increases. REM has the shortest leg movement duration and the longest interval time [9] [10]. <br><br>Therefore, people's sleep and wakefulness can be studied by using body movement characteristics. De CP et al. [11] extracted the characteristics of body motion signals and based on empirical judgments to better distinguish between wakefulness and sleep states, but the distinction between the various stages of sleep is more ambiguous. <br><br>In this article, the BCG signal and body motion signal are comprehensively selected, and the corresponding characteristic signals are extracted to determine the various stages of sleep. In view of the need to implement portable sleep quality diagnosis, a relatively small and effective algorithm is adopted to perform automatic sleep staging recognition. <br>After a comprehensive comparison, the improved BP neural network algorithm was selected for the study of sleep staging because of its low computational complexity, high feasibility, and strong robustness [12]. First of all, the neural network algorithm is designed using the sleep staging labels obtained from experiments. Subsequently, HRV and body movement feature values ​​are extracted according to the central impact signal and body movement signal of the sleep process, and the network training is combined with genetic algorithm. Finally, the trained network is used to stage sleep and compare with actual detection results. <br><br>Feature parameters and pattern recognition <br>Feature parameters <br>ballistocardiographic (BCG) is one of the most basic and important physiological signals of the human body. An instrument sensitive to pressure changes is used to detect the pressure change signal generated by a series of weak movements of the human body caused by the heartbeat and convert the pressure change signal into an electrical signal to record in a waveform, the waveform of which is a BCG signal [13]. The BCG signal waveforms are: F, G, H, I, J, K, L, M, N waves, as shown in Figure 1.

In addition to the performance of by-intersectional nerve activity, the physical activity of a person during sleep is much less than that of the awakening period, especially the periodic leg movement signal and the turn-over signal are more significant. Significant differences can be seen in the characteristics of periodic leg movements at different stages of sleep. Relative frequency, duration, and effects of awakening all decrease with the gradual deepening of sleep, while the interval between leg movements increases. REM has the shortest leg movement duration, while the longest interval is 9.10.<br><br>Therefore, people's sleep and awakening can be studied by using physical characteristics. De CP et al. 11 By extracting the characteristics of physical signal and judging based on experience, the better regions are divided into sobriety and sleep state, but the distinction between the various stages of sleep is more vague.<br><br>In this paper, the BCG signal and the physical signal are selected comprehensively, and the corresponding characteristic signal is extracted to judge the stages of sleep. In view of the need to achieve portable sleep quality diagnosis, a small and more effective algorithm is used for fully automatic sleep stage recognition.<br>After the comprehensive comparison, the improved BP neural network algorithm was selected to carry out the identification study of sleep stage because it has the characteristics of small computation, high feasibility and strong robustness. First, neural network algorithms are designed using experimental sleep stage labels. Subsequently, HRV and motion characteristic values are extracted according to the shock signal and movement signal of the sleep process center, and network training is combined with genetic algorithms. Finally, well-trained networks are used to stage sleep and compare it with actual test results.<br><br>Feature parameters and pattern recognition.<br> Feature parameters.<br>The heart shock signal (ballistocardiographic, BCG) is one of the most basic and important physiological signals in the human body. Use instruments that are sensitive to pressure changes to detect pressure change signals from a series of weak movements of the human body caused by a heartbeat and convert the pressure change signals into electrical signals recorded in waveforms, which are BCG signals. The BCG signal waveforms are: F, G, H, I, J, K, L, M, N waves, as shown in Figure 1.

In addition to the parasympathetic activity, the body movement during sleep is much less than that in the wake-up period, especially the periodic leg movement signal and turn over signal are more significant [8]. Significant differences can be seen in the characteristics of periodic leg movement in different sleep stages. The relative frequency, duration and arousal effect decreased with the deepening of sleep, while the interval time of leg movement increased. REM has the shortest leg movement duration and the longest interval time [9] [10].<br>Therefore, people's sleep and wakefulness can be studied by using the characteristics of body movement. De CP et al. [11] by extracting the characteristics of body movement signals and based on empirical judgment, they can better distinguish awake state from sleep state, but they can not distinguish each stage of sleep clearly.<br>In this paper, BCG signal and body movement signal are selected synthetically, and the corresponding characteristic signals are extracted to judge the sleep stages. In view of the need for portable sleep quality diagnosis, a less computational and more effective algorithm is adopted for automatic sleep stage recognition.<br>After comprehensive comparison, the improved BP neural network algorithm is selected to identify sleep stages, because of its small amount of calculation, high feasibility, strong robustness and other characteristics [12]. First of all, the neural network algorithm is designed by using the sleep stage tags obtained from experiments. Then, HRV and body motion eigenvalues are extracted from the central impulse signal and body motion signal during sleep, and the network training is combined with genetic algorithm. Finally, the trained network is used to stage sleep and compare with the actual detection results.<br>Feature parameters and pattern recognition<br>characteristic parameter <br>Cardiac shock signal (BCG) is one of the most basic and important physiological signals. The pressure change signal generated by a series of weak movements of the human body caused by the heartbeat is detected by the instrument sensitive to pressure change, and the pressure change signal is converted into electrical signal for waveform recording, and the waveform is BCG signal [13]. BCG signal waveforms include F, G, h, I, J, K, l, m, n waves, as shown in Figure 1.