A Sub-band Spectral Analysis for Electrocardiography

This paper addresses the adaptive version of a signal processing technique for power spectrum analysis of biomedical signals. The proposed algorithm combines the autoregressive (AR) model and the multirate system technique to divide the computation effort, the algorithm is based on an adaptive steepest descent algorithm to handle slow time drifting signals. The adaptive AR model provides an access to parameterized spectral representation, and the multirate technique partitions the signal into high frequency and low frequency sub-bands. The partitioning of the spectrum enables separate computation of the frequency sub-bands. The sub-band signal can then be concatenate to form a complete spectrum. The method than allows one to use smaller amount of memory to sequentially compute the frequency spectrum. The method is used on data from the MITBIH arrhythmia electrocardiography (ECG) database for verification.