Wavelet based reduction of stimulus artifact in evoked otoacoustic emissions testing

The initial segments of evoked otoacoustic emissions are contaminated by a stimulus artifact. This artifact inhibits examination of high frequency cochlear responses, which have short latencies. We propose an artifact reduction technique that draws on discrete wavelet transform pre-processing and neural network adaptive noise reduction methods. We show empirically that the proposed pre-processing method not only improves artifact reduction by an order of magnitude but also enhances signal-to-noise ratio in the response region