Abstract: A method of optimal input signal design for MIMO system identification is proposed. We first apply the wavelet multiresolution theory to the least squares MIMO system identification, and combine the input signal with the control performance index of the closed-loop system, then develop a method of optimal experiment design. Due to the multiresolution characteristic and the pre-whitening ability of the wavelet transform, this method could get better results than the existing system identification methods. The simulation result shows the effectiveness of the proposed method.