An integrated sleep stage classification device based on electrocardiograph signal

In this paper, a portable, easy to use, and real-time sleep stage classification device is presented. A simpler approach using raw features of ECG signals for sleep stage classification has been developed. Only one lead of ECG signal is required for operation which makes the device easily operable and only requiring user to attach 3 electrodes to the body. The device is constructed with singleboard computer and electronic circuits with Surface-Mount Device (SMD) technology making it small in size and highly portable. Classification can be done real time with an average delay of 20 seconds. Two sleep stages, namely the Awake and Non-Wake Sleep, can be differentiated by Random Forest algorithm. Data from the MITRA database was used for training and testing. Data from one patient was left out from the training set and used for testing. The device's recognition has high performance with weighted average value of 0.941 for Precision and value of 0.942 for Recall.