ST deviation and shape change are frequently used in ECG diagnosis. However, it is difficult to locate ST segment exactly, which makes the analysis of ST shape is not easy. Taking into account that the starting point of ST segment J is hard to locate, and the segment from S to the end of ST segment has a good differentiation, so we chose the segment from S to the end of ST segment as the ST segment. First, the ECG signals were analyzed by wavelets transform to identify the S and the ending point of the ST segment(the starting point of T) accurately. After we smoothed this segment, the ST shape was discriminated with improved dynamic time warping algorithm. The proposed method was tested with MIT/BIH European ST-T database, it performed well in recognizing five common depression types,upsloping, sinking, horizontal, sagging and arching types.
|