Design and implementation of an efficient Active Noise Control system

Abstract

Noise is an undesired, but unavoidable phenomenon. We cannot stop its generation at the source level, but control it at the listener level up to some extent either by passive or active method. In our work we are only concerned about the active method. In this paper we have proposed several types of adaptive algorithms for updating the weights of the digital filter, which acts as the controller. First we have proposed Filtered-X LMS algorithm which is very simple to implement and easy to understand. Then some amount of nonlinearity is introduced in the primary path and the performance is seen to be degraded. So this problem is sorted out by assuming a nonlinear controller using nonlinear algorithms like Volterra series method, Back propagation method for multi-layer perceptron, FLANN filter. After studying these algorithms we introduce a completely different type of algorithm known as evolutionary computing methods, which is based on the population based searching techniques. In this field we have studied 3 algorithms. i.e. Genetic Algorithm, Particle Swarm Optimization, Differential Evolution. A brief comparison is made between them and also the performance is studied in presence of nonlinearity.