Evaluation of Ballistocardiogram Based on Short-Time Energy and Dynamic Time Warping

Ballistocardiogram (BCG) has potential to monitor cardiovascular system in daily lives. Because artifacts caused by body movement are inevitable, developing evaluation methods to distinguish contamination is necessary. So far, there have been methods based on template or information criteria for assessing periodicity or complexity of raw signals. These methods rely on another reference signal to build template or priori knowledge about the raw signals, which are not always available. In this paper, we present a novel BCG evaluation method based on short-time energy for extracting feature, and dynamic time warping for calculating the similarity of extracted features, considering the temporal variation of BCG. Experimental results show that without template or other reference signals, proposed method can distinguish the resting state, small range motion state and wide range motion state. This greatly improves the reliability of BCG quality evaluation, and therefore the potential of BCG cardiovascular system monitoring.