Signal segmentation using time-scale signal analysis

Change-point detection is a fundamental problem in many areas of signal segmentation, feature extraction and classification. There are various statistical methods that can be applied to estimate boundaries of signal segments. The paper presents another method based upon the wavelet transform and signal decomposition used to detect signal irregularities. This approach can be very efficient if the initial wavelet function is carefully chosen. Basic methods presented in the paper are verified for simulated sequences at first and then used for biomedical signal segmentation. Resulting algorithms are written in the MATLAB environment.