Abstract: To extract visual evoked potential (VEP) from strong electro-encephalogram (EEG) background activity, the internal model filtering is firstly performed in terms of the internal model parameters of actual VEP (obtained from the superposition of several EEG records) estimated in an on-line manner via the recursive least squares (RLS) method, then the wavelet transform technique is combined with the internal model filter developed previously. The real clinical applications have shown more desirable result by using the combined approach than by the typical VEP-based internal model filtering approach or the pure wavelet transform approach separately.