Bio-signal Characteristics Detection Utilizing Frequency Ordered Wavelet Packets

An application of wavelet packet is presented for electrocardiogram (ECG) and magnetic resonance spectroscopy (MRS) characteristics detection in this study. A fully automated system is developed to detect the "R" peaks which are beat designators and are used consequently to locate other ECG characteristics. They include "P", "Q", "S" and "T" waves along with "ST" segment shift. The peaks and the area under the peaks of MRS signals are also detected. The Daubechies wavelets are selected as base processing filters. Frequency ordered wavelet packets (FOWPT) is utilized to generate a time-frequency plot of the signal used for further processing. The algorithm is validated on MIT-BIH database. The proposed beat detector achieved sensitivity of 99.18%plusmn2.75 and a positive predictivity of 98.00%plusmn4.45. The "P" wave detector achieved sensitivity of 51.69% and a positive predictivity of 53.64%.