Least squares time sequenced adaptive filtering for the detection of fragmented micropotentials

The authors show a signal enhancement technique based on a time-sequenced adaptive filter (TSAF). The time-sequenced approach allows enhancement of signals with rapidly time-varying statistics if this time-variance exhibits cyclostationary characteristics. In event based signals such as ECGs this time-invariance is synchronised with each beat and is hence cyclostationary. Instead of conventional LMS based TSAF the authors are controlling their filter through a least squares (LS) updating criterion. They are able to show a significant convergence rate improvement using LS adaption and therefore they require much shorter data sets.<>