Waveform Processing of Electrocardiogram with Neural Network and Non-contact Measurement using Kinect for Driver Evaluation

Abstract

In recent years, the measurement method of biological information using non-contact sensor have become popular. And, in-vehicle measurement systems for a driver evaluation are also needed in order to decrease the serious traffic accidents. In this research, improvement method of heart rate measurement with high accuracy by using the non-contact sensor “Kinect” is examined. As the reliable biological information, the Holter monitor of electrocardiograph is combination used for measurement experiment. Moreover, to extract high level features from electrocardiogram, a neural network based autoencoder has also constructed. In this paper, about the constructed neural network, the effectiveness of abnormal R wave detection through learning for waveform characteristics of electrocardiogram and the possibility of application for reconstruction of the heart rate signal from Kinect are discussed.