Independent Component Analysis-Based Feed-Forward Neural Network for Applications of Active Noise Control
碩士 === 中興大學 === 機械工程學系所 === 95 === In this study, an application of active noise control (ANC) using an independent component analysis-based feed-forward neural network (FFNN) is investigated. We consider a speech source and a noise source generating mixture signals in the space. Two microphones loc...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2007
|
Online Access: | http://ndltd.ncl.edu.tw/handle/73680378043527749812 |
Summary: | 碩士 === 中興大學 === 機械工程學系所 === 95 === In this study, an application of active noise control (ANC) using an independent component analysis-based feed-forward neural network (FFNN) is investigated. We consider a speech source and a noise source generating mixture signals in the space. Two microphones located at two distinct places are used to measure the mixture signals. Since the signal sources and the transmission paths are unknown, we apply a FFNN with MJH algorithm(FFNN_MJH)for the measured signals to obtain estimates of the signal sources. These estimates are then used for an ANC controller such that suitable control signal can be generated to drive a loudspeaker. It is desired that one of the microphone can observe the speech while ignoring the noise by use of the loudspeaker. Computer simulation shows that the observed microphone can effectively retain the speech while attenuating the noise with respective to tonal or harmonic noises. The proposed system also maintains a certain degree of robustness with respective to the uncertainty in the transmission paths of the loudspeaker, demonstrating its feasibility.
|
---|