Research on Adaptive Algorithm for Weight Partial Update of ANC
Shifei Hui,Mingjiang Wang,Yufei Han
DOI: https://doi.org/10.1109/ICICSP48821.2019.8958571
2019-01-01
Abstract:In the process of driving, people pay more and more attention to the vehicle noise. The ANC (Active noise control) system makes up for the shortcomings of the noise reduction system designed by analog circuit in the application of automobile noise reduction. When the noise environment changes, it can change the parameters of the filter in real time to achieve the ideal effect of noise reduction. However, the closure of interior space, the reflection of sound waves and the diversity of noise propagation paths seriously affect the noise reduction effect of ANC system. In order to eliminate them, the length of the weight filter needs to be increased. After the filter length is increased, the noise reduction effect of the system plays a positive role [1]. However, it also brings the increase of system complexity and computation. For these reasons, this paper proposes a weight partial update algorithm to reduce the computational load. Although partial update algorithm reduces the computational load, it also affects the rate of convergence [2]. Therefore, this paper proposes a new variable step algorithm to improve it. To sum up, in order to improve the effect of noise reduction ANC algorithm, to reduce the amount of calculation and improve the convergence speed, we need a new algorithm. Basing on the traditional LMS algorithm, a new weight partial updating algorithm and variable step size algorithm is proposed in this paper. The computational complexity, convergence speed and noise reduction effect of the algorithm are compared with the two algorithms that has better noise reduction effect and appropriate rate of convergence [3]. Partial updating algorithms mainly include continuous local iterative algorithm, random local iterative algorithm, Mmax local iterative algorithm, and selective local iterative algorithm. Continuous local iteration algorithm and random local update algorithm can not complete the filter convergence quickly. So we choose selective local iterative algorithm to complete the algorithm design.