Design of discrete coefficient FIR filters using LMS algorithm

Two design techniques are proposed that can be used to design high-order discrete coefficient (FIR finite impulse response) filters. Both techniques are stochastic, in the sense that they randomly search in the discrete space of the filter coefficients for more optimal solutions than an existing one. The LMS algorithm is used to confine the search to those points that have a higher chance of being a good solution. The unique feature of the proposed techniques is their computing time, which increases only proportionally to the square of the filter length, as opposed to the other existing techniques, whose computer time increases exponentially with the length of the filter