Adaptive Real-Time Wavelet Detection

Detection of wavelets in noisy signals is important for many biomedical as well as other applications. An adaptive wavelet detection system is described, based on the cross correlation method. The method is restricted to finite-length, nonoverlapping wavelets. A random reference correlator (RRC) circuit is used to estimate the correlation between signal and template, eliminating the need for expensive hardware or time-consuming software correlation calculations. The adaptation of a template (from an initial estimate or to track slow wavelet variations) and its signal-to-noise ratio are investigated both theoretically and by simulation. Synthesized and real ECG signals are used to demonstrate the performance of the system. The algorithm for ECG processing was implemented in real time on a relatively slow microprocessor (6501 with 1 MHz clock). Detection results are shown to be compatible with regular digital correlators and relatively insensitive to baseline shifts and pulsative interferences.