tailieunhanh - 32 Inverse Problems in Microphone Arrays

An acoustic enclosure usually reduces the intelligibility of the speech transmitted through it because the transmission path is not ideal. | A. C. Surendran. Inverse Problems in Microphone Arrays. 2000 CRC Press LLC. http . Inverse Problems in Microphone Arrays . Surendran Bell Laboratories Lucent Technologies Introduction Dereverberation Using Microphone Arrays Simple Delay-and-Sum Beamformers A Brief Look at Adaptive Arrays Constrained Adaptive Beamforming Formulated as an Inverse Problem Multiple Beamforming Matched Filtering Diophantine Inverse Filtering Using the Multiple Input-Output MINT Model Results Speaker Identification Summary References Introduction Dereverberation Using Microphone Arrays An acoustic enclosure usually reduces the intelligibility of the speech transmitted through it because the transmission path is not ideal. Apart from the direct signal from the source the sound is also reflected off one or more surfaces usually walls before reaching the receiver. The resulting signal can be viewed as the output ofa convolution in the time domain of the speech signal and the room impulse response. This phenomenon affects the quality of the transmitted sound in important applications such as teleconferencing cellular telephony and automatic voice activated systems speaker and speech recognizers . Room reverberation can be perceptually separated into two broad classes. Early room echoes are manifested as irregularities or ripples in the amplitude spectrum. This effect dominates in small rooms typically offices. Long-term reverberation is typically exhibited as an echo tail following the direct sound 1 . If the transfer function G z of the system is known it might be possible to remove the deleterious multi-path effects by inverse filtering the output using a filter H z where 1 H z D Typically G z is the transform of the impulse response of the room g n . In general the transfer function of a reverberant environment is a non-minimum phase function . all the zeros of the function do not necessarily lie inside z 1. A minimum .