Study On Sleep Staging Method Based On HRV Analysis

Sleep is a necessary process of life and it is an important part of body recovery, integrate and memories consolidation. Sleep classification is the basis of sleep research and it is precondition of sleep quality assessment. Therefore, it is of great significance to realize sleep monitoring and sleep quality evaluation under home environment and provide vital parameters for sleep apnea screening. Sleep staging is an important basis of sleep quality evaluation and related disease diagnosis.Studies have shown that heart rate variability presents a cyclical change in sleep, which is similar to EEG. Therefore, heart rate variability and sleep has the inseparable relationship. This paper provides a new method of sleep stages classification, which uses RR intervals to identify different sleep stages, laying the foundation that it can be done at home. RR intervals can be taken advantage to classify different sleep stages by experiment. This is a supplement of the traditional way which is rely on EEG alone. The content of the research is as follows:1.The correlation of heart rate variability and sleep stages is analyzed in this paper. Firstly, we demonstrate the correlation between HRV and sleep stages from Time Domain including scatterplot, standard deviation and time domain chart aspect. Secondly, we carry on Fast Fourier Transform to give the demonstration from the angle of Frequency domain.2.Principle component analysis(PCA) is used in this paper to reduce redundancy of raw data. In order to get a more comprehensive and accurate result, we usually needs to extract many characteristics of multiple indicators. But the redundancy and correlations between different indicators will impact the results. Therefore, data dimension reduction can eliminate collinearity among variables, decrease the number of variables and replace the original interrelated data by less unrelated principal components. Synthesis and simplify original data solve the problem of model excessive fitting according to the contribution of principal components.3. Support vector machine(SVM) solves such practical problems as small sample, nonlinear and high dimension according to its theoretical system of the machine learning for small sample situation. This paper uses improved grid search method to seek the optimal parameters. We can get the best parameters by comparing and analyzing the result of the simulation through different steps and different range. And it improves the accuracy rate and reliability of this kind of classification at the same time.This paper realizes the sleep stage pattern recognition method based on HRV. We receive 80% accuracy by PCA dimensionality reduction pretreatment and SVM sleep pattern recognition. At the same time the sleep pattern recognition method has better antinoise ability according to the result obtained by simulation analysis, which can satisfy clinical needs.