Comparison of Linear and Non-linear Feedforward Algorithms to Control Chaotic and Impulsive Noise

Linear feed-forward adaptive algorithms are widely used for creating a quiet zone inside an enclosure for different noise signals like tonal, multi-tonal, broadband, impulsive and chaotic. However, these control strategies do not give optimal results under the presence of non-linearities. Non-linearities can occur in an active noise control system due to the following reasons: (a) Non-linear noise like noise generated due to vacuum cleaner, compressor, mixer and grinder, (b) Non-linear behavior of actuators and sensors and (c) A very high sound pressure level of noise, e.g., noise generated during rocket launch. Conventional ANC control strategies, like FxLMS algorithm, give poor performance and even sometimes fail to control this type of noise signal. Therefore, mitigation of these non-linear noise signal, in which adaptive algorithms give better results for creating a quiet zone inside the enclosure, need to be identified. In this study, we explore linear and non-linear feedforward control strategies to control chaotic and impulsive noise. Linear algorithms studied in this paper are (a) Filtered-x least mean square (FxLMS) and (b) Filtered-x recursive least square (FxRLS), and non-linear algorithm is (a) Filtered-x back propagation neural network (FxBPNN). The performance of these algorithms in terms of convergence time, steady state noise, stability (or robustness) and computational complexity to reduce chaotic and impulsive noise in a vibro-acoustic cavity is discussed.