Effective Extraction and Filtering of Frequency Components in Physiological Signals Using Sum-of-Sinusoids Modelling

In biological signal processing, modelling and extraction of specific frequency components constitute an important procedure for filtering signal components of interest as well as artefact removal. Under some interference scenarios, a satisfactory elimination of artefacts from the signal must be even performed by subtraction of an artefact waveform model or template, rather than the use of linear band-pass filters. That is the case of the gradient artefact induced in the EEG within the fMRI scanner, which cannot be characterized by a specific bandwidth or spectral content. This paper presents a simple and accurate approach based upon sum-of-sinusoids modelling for signal and artefact frequency components representation in physiological signals. According to the proposed method, each signal frequency component is approximated as a sinusoid, whose amplitude and phase parameters are estimated by making use of the Discrete Fourier Transform (DFT). The proposed approach reveals to perform an effective modelling and extraction of ECG signal components as well as underlying gradient artefacts in the EEG signal