Blind channel identification exploits the measurable channel output signal and some prior knowledge of the statistics of the channel input signal. However, in many scenarios, more side information is available. In digital communication systems, the pulse-shaping filter in the transmitter and the anti-aliasing filter in the receiver are often known to the receiver. Eaploitation of this prior knowledge can simplify the channel identification problem. In this paper, we pose the multipath identification problem as solving a group of linear equations. While we solve the linear equations in the least-square meaning, a weight matrix can be introduced to improve the performance of the estimator. The optimal weight matrix is derived. Compared with the eristing Linear Prediction ( LP ) based multipath identification approach, the proposed approach offers a substantial performance gain.