The present paper describes an adaptive signal enhancer (ASE) method for improving signal-to-noise ratio (SNR) and also for tracking the variation of brain-stem auditory evoked potential (BAEP). The enhancer has two inputs: a primary input which is the original data to be processed, consisting of signal badly corrupted by noise, and a reference input. A method called dynamic averaging (DA) is introduced to obtain the reference input. The improved reference signal allows closer tracking of evoked potential in each subsequent trial. For BAEP measurement in human, the SNR is typically very low. It requires about 200 ensembles to generate the reference input acceptable for ASE, and tracking of the variation of BAEP can be obtained satisfactorily. Our results show that while ensemble averaging is still needed, the total number of ensembles is greatly reduced. The adaptive method thus reduces the time of measurement for BAEP compared with the conventional method of ensemble averaging.