Hardware Implementation of Fast Transversal Least Mean Square Algorithm in Acoustics, Speech, and Signal Processing (ASSP) Using TMS320C5X Processor

This paper describes the Normalized Least Mean Square algorithm and Fast Transversal Least Mean Square Algorithm for effective noise cancellation. The Simulink model for NLMS and Fast Transversal LMS Algorithm had been designed which results in a noise free signal as output in ASSP. The filter used here is adaptive filter and the algorithm used is Least Mean Square algorithm and Fast Transversal Least Mean Square Algorithm. The input given is the original speech signal/sinusoidal varying input where in white Gaussian noise / Random noise are deliberately introduced to the block. By varying the adaptive step size, Signal to Noise Ratio is determined and are compared for both the algorithms. Based on these results the optimum step size is found for noise free output and the best efficient algorithm is identified. The Fast Transversal LMS algorithm is found to be a suitable solution for adaptive filtering applications and hence chosen for implementation in hardware using TMS320C5X processor. Thus hardware has been implemented for effective removal of noise in audio and speech processing and it can be widely used in the detection of Narrow band signals in Broad band Noise.