Training in ABR
Calculating the residual noise in an ABR waveform
Description
This paper describes our early attempt to estimate the signal-to noise ratio of the averaged ABR. In the clinic, ABRs are recovered from the on-going background noise by averaging a number of sweeps. Normally, a test protocol will prescribe a fixed number of sweeps to be averaged and will recommend replications to be obtained. However, since both the ABR and the background noise differ across individual subjects both in magnitude and in other characteristics, such a test protocol can never ensure a given minimum ‘quality’ or signal-to-noise (response-to-noise) ratio, SNR, of the final recovered ABR. Therefore a statistical method is developed in order to estimate the SNR of the recorded ABR during the on-going averaging process. The method calculates the FSP, which is the squared ratio of the estimated magnitude of the ABR to that of the averaged background noise. The method can be employed on-line as an adaptive strategy (1) to estimate the number of sweeps necessary to obtain a given minimum SNR (quality) of the ABR recorded at supra threshold levels, or (2) to automatically detect the presence of an ABR near threshold.
