Effective Extraction of Visual Event-Related Pattern by Combining Template Matching With Ensemble Empirical Mode Decomposition

In the field of signal processing, it is always a major challenge to extract event-based weak or low signal in the presence of high background noise. Conventionally, this is achieved by trigger-based averaging, which suppresses uncorrelated background noise and unmasks the event related pattern. In some of the previouspapers, extraction of weak event related pattern is also achieved by decomposing the signal into a set of predefined basis functions, such as wavelets. We present here, a novel approach by combining template matching with the ensemble empirical mode decomposition (EEMD). The EEMD technique is applied to decompose the noisy data corresponding to single-trial event related potentials into the so-called intrinsic mode functions (IMFs). These functions are of the same length and in the same time domain as the original signal. Therefore, the EEMD technique preserves varying frequency content along the time axis. The effective extraction of the event-related pattern proposed in this paper relies on the elimination of IMFs, which capture the features corresponding to artifacts and brain signals, based on cross-correlation with a suitable template extracted from the evoked potential obtained by the conventional unrestricted averaging across a large number of trials. We illustrate the method and compare it with conventionally used single channel wavelet-based approach for denoising visual evoked potentials during the measurement of visual evoked electroencephalogram response.