Blind biosignal classification framework based on DTW algorithm

Biosignal is a noninvasive measurement of the status of internal organism, such as electrocardiogram (ECG), electroencephalogram (EEG), and electromyogram (EMG), etc. With machine learning techniques, these biosignals are normally classified into one of a number of disease categories. Hence, they are ideally suited to support clinician in making diagnostic decision. However, if a given biosignal is an unknown type, none of the existing classification algorithms can be considered workable. In this paper, an intelligent framework that is able to automatically identify ECG from an unknown biosignal is described. In which, the first phase of the research is illustrated in detail, which focuses on classifying an unknown biosignal into ECG or other categories, by employing dynamic time warping (DTW), combined with clustering algorithm. The proposed framework consists of two major components: biosignal template construction and classification process. Biosignal template construction includes biosignal acquisition and segmentation, template optimization and management; while the classification process involves several sub-processes: biosignal preprocessing, biosignal pattern matching and majority voting. The experimental results demonstrate the effectiveness of the framework as well as the classification methodology.