Sleep staging methods using pulse frequency variability

A novel sleep stage classification method using pulse rate variability analysis was proposed in order to learn one's sleep pattern in home environment. First, blood oxygen fingerstall was employed to obtain the original pulse data; the data was processed with multi-distinguishability wavelet transformation method; the first layer of the signal was extracted to obtain the pulse interval with a self-adapted sliding window algorithm. Next, the pulse rate variability features were extracted from the pulse interval information acquired before. Finally, the sleep stage could be classified according to the differences between various stages of sleep. During the classification process, time domain, frequency domain as well as nonlinear domain of the pulse rate variability signals were main references. After all the features being extracted from the data, the features were estimated and the sleep staging classification was given out. The experiment results were evaluated and analyzed. In this experiment environment, the precision of sleep stage classification can reach 76%, indicating the effectiveness of our method, which also means that the method can classify one's sleep stage efficiently in home environment with low cost.