A low-cost multichannel active noise control system for personal quietude

Abstract

This paper describes the development and performance of a compact, low-cost, one-input, two-output, two-error multichannel active noise system for personal (small-region) quietude. Operation of the system encompasses both modeling and attenuation phases and includes automatic restart capabilities if performance is lost. To achieve a robust single DSP implementation, several modifications of the filtered-X LMS adaptive controller have been employed, including partial gradient calculation, sign-error adaptation, and approximate iterative data normalization of the algorithm step size. Acoustical factors have been addressed in the selection of filter lengths, sampling rate, and convergence factors, as well as in the placement and selection of sensors and output sources. Experiments indicate that the system can achieve 20 dB of noise reduction in the frequency band between 100 Hz and 1700 Hz for localized random noise sources.