Active Noise Control Over Spatial Regions

Abstract

This thesis investigates active noise control over a large spatial region using efficient control systems. Active noise control (ANC) utilises secondary sound sources to cancel primary noise based on the principle of destructive interference, and has the advantage of high flexibility and easy adaptability. ANC over a large spatial region (spatial ANC), which requires multiple sensors and multiple secondary sources in the system, creates a large-sized quiet zone for multiple listeners in three-dimensional spaces. The existing multichannel approaches are not very efficient in spatial ANC, as the noise cancellation is optimized only around the error sensors. This thesis provides new adaptive solutions for spatial ANC in general noise fields and optimal methods for spatial ANC in sparse noise fields.

In terms of adaptive solutions for spatial ANC in a general noise field, our approach is to utilize the wave-domain signal processing technique. Several outcomes resulting from this approach are (1) the design of the feedback wave-domain ANC system, and derivation of the filtered-x least mean square wave-domain approaches; (2) systematical formulation of the wave-domain ANC into different minimization problems and different updating variables, and derivation of four normalized wave-domain approaches. We show that, compared to the conventional multichannel approaches, the proposed wave-domain ANC approaches can achieve significant noise reduction over the entire spatial region with faster convergence speed.

In terms of the optimal methods for spatial ANC in a sparse noise field, our approach is to incorporate the l1-norm constraint from compressive sensing into the spatial ANC. Several outcomes resulting from this approach are (1) derivation of the l1-constrained multichannel approaches; (2) derivation of the l1-constrained wave-domain approach. We show that, compared to the conventional multichannel approaches, the proposed l1-norm constrained approaches can reduce the number of active secondary sources with faster convergence speed.

In addition, this thesis investigates the best possible spatial ANC performance for a given system, by analyzing the signal space spanned by the secondary sources within a given acoustic environment. The proposed subspace method can obtain best possible ANC performance and is demonstrated to be a feasible solution, especially when the secondary sources are not sufficient to cover all orthogonal spatial modes according to the spherical harmonic theory.