Design of Portable EEG Acquisition System and Sleep Staging Method Based on Multi-scale Entropy Feature

In this paper, we try to design of portable EEG acquisition system and sleep staging method based on multi-scale entropy feature. Portable EEG signal acquisition equipment adopts
$\mathbf{ARM}+\mathbf{ADS}\mathbf{1299}$
scheme, ARM processor adopts the STM32F103 series single-chip microcomputer, and TI ADS1299 as the signal acquisition front-end analog chip. We collected EEG signals by a neighborhood comparison digital filtering algorithm and a band-pass filter, which removes the pulse interference of the EEG signals, the artifacts of EMG and OMG. The pure EEG signal eas extracted by time domain, time-frequency domain, and nonlinear feature, and 10 features of the EEG signal segment every 30s are obtained, including 5-time domain features and 4 energy features based on wavelet packet transform. Multi-scale entropy MSE, sleep EEG dataset was expanded through Sleep-EDF, and feature extraction was performed. Random forest and SVM were used to classify and extracted features. Staging results show that the classification accuracy of random forest is higher than that of SVM, which accuracy rate can reach 92.5%, and the Kappa value above 0.85. To verify the reliability of our system, we take some clinical experiments in Jinniu Hospital of Sichuan Provincial People's Hospital.