Neural network implementation of Least-Mean-Square adaptive noise cancellation

Adaptive systems have been used in a wide range of applications for almost four decades. It includes adaptive equalization, adaptive noise cancellation, adaptive system Identification, Inverse modeling, linear prediction etc. It is often assume noise to be random process. This paper describes the concept of neural network implementation of adaptive noise cancelling using Least-Mean Square adaptive filter algorithm. In this, the coefficients are adjusted by an analog neural network instead of numerical adaptive algorithms. Due to its real time processing capabilities, the neural network can optimize the coefficients of the adaptive filter at each new received sample, which is especially useful in non-stationary environments. Due to the parallel and analog nature of the processing, the time needed by the neural network for computation of those coefficients is small. Compared to LMS-ANC direct method, the mean-square error in LMS using neural network can be changed. Simulation results demonstrate satisfactory performance.