Subspace Averaging Of Auditory Evoked Potentials

The auditory evoked potential (AEP) is an electric potential generated in the brain in response to auditory stimuli. It has clinical importance in the detection of newborn infant hearing loss, among other applications. The signal-to-noise ratio (SNR) of the AEP is low, so signal averaging is typically employed to estimate it. Often, thousands of trials must be averaged before a sufficiently high SNR estimate is obtained. In this research, we have developed an AEP averaging method called subspace averaging. The subspace averaging method projects onto the signal subspace: the span of the principal eigenvectors of the signal autocorrelation matrix. The signal subspace has low dimensionality and captures the key features of the signal. Also, we introduce a new SNR estimator for AEP trials. Using our estimator, we compare SNR estimates of conventional averaging and subspace averaging. The subspace average has higher SNR compared to the conventional average.