Information Processing Models For Sleep Staging Based On Intelligent Mattress

With the development of social economy and science and technology in our country,there's been a big breakthrough in modern medical theory.Nowadays,instead of receiving medical care,people are more concerned to stay healthy.The new concept of active health care mode includes detecting early sings of illness,making a scientific assessment,adjusting body condition and promoting good health.Sleep Health Monitoring mattress as a low power consumption,non-intrusive and unobtrusive real-time health monitoring instruments,carry out the active health care ideal perfectly.At present under the condition of medical treatment shortage the Sleep Health Monitoring mattress will develop into a home sleep monitoring method to assist hospital to screen patients in the future.Sleep Health Monitoring mattress is a kind of signal acquisition device.In this article we introduce a high performance system to process the mixed-signal collected by the piezoelectric ceramic sensor embedded mattress.The mixed signals we collect contains the activity signals of the user,such as heartbeat component,respiration signal and lots of noise like body movements,ECG artifact.Need to find useful informational in the mixed signals,among them the Ballistocardiogram(BCG)signal can reflect the heart rate and the ECG-Derived Respiration(EDR)can extract respiration rate.But it's difficult to extract useful component,the most important means to obtain target signal is the noise suppression.In this paper firstly analysis the spectrum structure of the mattress mixed signal by the characteristic of Spectral Centroid.Then build up an new model to classify the quality of sampled signals based on Wavelet Energy Entropy and decision tree.We divide the quality of signals into four categories:good signals,abnormal signals,body movement signals and off bed signals.According to the quality of each windowing signal,an intelligent signal processing method with the scenario analysis is proposed.Therefore the algorithm's adaptability and the accuracy of the results are both increased.Applying the signal processing method,the heart rate and the EDR can be extracted form the raw data.By these parameters,the Heart Rate Variability(HRV)and Cardiopulmonary Coupling(CPC)are calculated.Next define the public lab of Cyclic Alternating Pattern(CAP)and Polysomnography(PSG).It is verified that the PSG and CAP reflected in the new lab are relevant and the differences are not statistically significant.Transforming the characteristics of CPC and HRV into the feature images and send them to the VGG16 which is a kind of classical network mode training sleep stage classification rules.The test results show that our system increases the diagnosis result reliability and accuracy,enhances the diagnosis anti-jamming capability.This system provides a new idea for sleep quality assessment.