Abstract: Channel estimation is a fundamental concern in wireless communication.An improved semi-blind channel estimation method using parametric multipath channel modeling is proposed,which is applied to static multipath channel.This method derives cost function based upon sub-channel response matching(SRM) criterion.To begin with,a global minimum is located through linear searches and is regarded as an initial estimated vector.Secondly,Steepest Descent Approach (SDA) is exploited to refine the initial estimated one.Computer simulations show that the proposed SDA is superior to Gauss-Newton(G-N) method.