Optimisation of frequency-band adaptive filtering in real-time beat-by-beat high-gain ECG processing

Abstract

Although beat-by-beat processing can allow the recovery of transient microvolt signals that are impossible to recover with time-domain averaging, the computational implications can be significant. Frequency-band adaptive filtering is a beat-by-beat technique which illustrates this problem. In particular, the real-time processing required by such approaches may be beyond the capabilities of a single modest DSP processor. Following an investigation of the spectral power and noise characteristics of segmented ECG signals, an optimised multi-frequency-band adaptive filter has been implemented which has reduced the computational demands and is successful in the registration of ventricular late potentials in noisy circumstances on a beat-by-beat basis.<>