Use of Transform Domain Least Mean Square Algorithm in Active Noise Control for Music Noise Reduction

摘要

This study focuses on the use of the Transform Domain Least Mean Square (TDLMS) algorithm to reduce music noise from outdoor concerts. Outdoor concerts are ambivalent noise sources because concert sounds please the audience, while irritating nearby residents. Active noise control based on the Least Mean Square (LMS) algorithm can be an applicable method for noise reduction. When music noise serves as an input signal in the algorithm, the statistical characteristics of music noise can degrade the convergence rate of the LMS algorithm. To solve the problem, TDLMS is applied to compensate for the degraded convergence behavior caused by the music noise. Discrete Cosine Transform (DCT) is selected as a fixed orthogonal transform of TDLMS algorithm. An outdoor experiment with single-channel active noise control is conducted to show the improvement of the convergence behavior between Filtered-XLMS and TDLMS. We analyze the convergence rate of each algorithm with Root Mean Square (RMS) between 100 [Hz] to 1 [kHz] as a performance index. Improvement of convergence rate with TDLMS is validated in three different music samples. The experiment result shows that the TDLMS converges faster than the Filtered-X LMS in music noise control.