Detection and correction of outliers in foetal heart rate time series

When subjected to poor quality signals, foetal heart rate (FHR) estimation algorithms produce random erroneous estimates. An adaptive filter configured as a forward/backward predictor corrector (FBPC), is implemented to detect the identifiable erroneous estimates (outliers) and to provide corrections. The FBPC structure proposed is based on the least mean square (LMS) adaptive filter whose coefficients track the underlying characteristics of the FHR signal. The effectiveness of this approach is demonstrated using heart rate time series data. Good performance has been achieved even during epochs with low quality signals.