A Discrete Fourier Transform method for alignment of visual evoked potentials

In this paper, we consider alignment of visual evoked potentials (EP) in the Discrete Fourier Transform (DFT) domain. Visual EPs have important clues for diagnosing medical problems such as multiple sclerosis and optic neuritis. The amplitude of visual EPs are usually smaller than the amplitude of spontaneous EPs which causes difficulties in reliably finding the latencies and amplitudes of important positive and negative peaks in the evoked responses. Therefore, noise cancellation becomes important for determining the features of interest in these waveforms. A well-known noise cancellation method is averaging multiple evoked potentials. Averaging after alignment of EP waveforms can improve the waveform quality substantially since usually evoked potentials have different characteristics and therefore have different latencies and amplitudes in response to the same visual stimulus. In this paper, we use a time alignment method which simultaneously reduces the spectral differences between all waveforms by minimizing the linearly phase shifted forms of the DFTs of these waveforms. We demonstrate that this method successfully aligns multiple visual EPs and achieves a smooth averaged waveform with reduced noise.