Evoked potentials (EPs) are time-varying signals typically buried in relatively large background noise. Tracking variations in both the latency and the amplitude of EPs are important in quantifying properties of the nervous system. Adaptive filtering is a powerful tool for tracking such variations. In this article, EP signal estimation by adaptive filtering was introduced in the framework of an adaptive signal enhancer (ASE). Two inputs exist in ASE: the primary input, or the raw recorded data, that contains a signal plus noise, and the reference input that contains a signal related to, but not necessarily having the same waveform as the signal components in the primary input, and an additive noise, which is not correlated to the noise in the primary input. ASE requires a proper reference input signal for its optimal performance. Several effective reference input designs are introduced in this article. The results confirm the superior performance of ASE over the traditional ensemble average method.