De-noising of auditory brainstem response via diffusion and wavelet transform

Evoked Potentials are event-related activities that occurred as an electrical response from the brain to different sensory stimulations of nervous tissues. In this paper, auditory evoked potentials (AEP) brain responses were collected and examined. The data collection was done twice with three different levels of sound and frequencies. The auditory brain response data were extracted from the noisy original data using the averaging technique and set as a reference signal. We propose new approaches for feature extraction of the auditory brain response using wavelet transforms and diffusion filters algorithms. The wavelet transform has the ability to resolve the data into various levels of decomposition, which facilitate its representation in the frequency and time domain. The diffusion filters, on the other hand enhanced the extracted signals resulting in the noise suppression and thus reducing the error. Performance analysis was done based on signal-to-noise ratio (SNR), mean squared error (MSE) and peak-signal-to-noise ratio (PSNR). The outcome shows that the diffusion technique produces better performance than wavelet transform in all the cases studied.