期刊
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
卷 28, 期 -, 页码 1479-1492出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TASLP.2020.2989582
关键词
Training; Optimized production technology; Frequency control; Feature extraction; Noise reduction; Speech processing; Signal processing algorithms; Active noise control; selective fixed-filter active noise control; frequency-domain filtering; filtered-x least mean square
资金
- Singapore Ministry of National Development
- National Research Foundation, Prime Minister's Office
- Cities of Tomorrow (CoT) Research Programme (CoT Award) [COT-V4-2019-1]
Conventional real-time active noise control (ANC) usually employs the adaptive filtered-x least mean square (FxLMS) algorithm to approach optimum coefficients for the control filter. However, lengthy training is usually required, and the perceived noise reduction is not immediately realized. Motivated by the practical implementation, we propose a selective fixed-filter active noise control (SFANC) algorithm, which selects a pretrained control filter to attenuate the detected primary noise rapidly. On top of improved robustness, the complexity analysis reveals that SFANC appears to be more efficient. The SFANC algorithm chooses the most suitable control filter based on the frequency-band-match approach implemented in a partitioned frequency-domain filter. Through simulations, SFANC is shown to exhibit a satisfactory response time and steady-state noise reduction performance, even for time-varying noise and real non-stationary disturbance.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据