Sleeping State Detection System Based On Smart Mattress

The monitoring of human health status during sleep is one of the research hotspots in the field of smart medical care in recent years.Since using the existing monitoring devices need close-to-body,which directly affects the comfort of the user during sleep.This paper proposes a non-contact smart mattress-based sleep detection system,which collects the pressure data of the human body perceived by the mattress during sleep,and combines intelligent signal processing algorithms to obtain human health information in real time.In view of the different biological characteristics of the human body,this paper proposes the following three detection schemes:(1)A rule-based and convolutional neural network based sleep motion recognition scheme is proposed here.The rule-based sleep motion recognition algorithm analyzes the received data using the variance and the threshold according to the characteristics of the human motion,and then recognizes the motion.The latter transforms the sleep motion recognition problem into a classification problem in machine learning,and uses the convolutional neural network to classify the data from the sensor array to realize sleep motion recognition.(2)A monitoring scheme for the respiratory rate of the human body is proposed here.Considering that the data collected by the sensor array come from some useless channels,thus the main channel selection algorithm is first used to obtain the signal with the selected appropriate channel,and then the obtained data is preprocessed by the smoothing algorithm and the baseline drift elimination algorithm to eliminate the noise interference and trend items.Finally,according to the regularity of respiratory movement,this paper selects the appropriate number of wavelets according to the respiratory frequency to reconstruct the respiratory waveform and calculate the respiratory rate of the user by the extreme point counting method.(3)A monitoring scheme of human heart rate signal based on multi-wavelet reconstruction and long-term and short-term memory neural network is proposed here.Based on the regularity and weakness of the heartbeat signal,this paper uses the multiwavelet reconstruction signal combination model to restore the heart rate signal,avoids the shape distortion of the single wavelet reconstructed wave,and optimizes the monitoring waveform.At the same time,according to the time series characteristics of the mattress data,this paper proposes two long and short memory networks,long vector and long sequence.The long sequence network is selected by experiment to extract the heartbeat waveform.Finally,the experimental data of each program was compared with the professional medical device data.The results show that the proposed sleep detection system can accurately obtain the physiological state information of the human body,and fully meet the needs of daily life monitoring sleep.