A New Robust Adaptation Mode Controller based on a Normalized Cross-Correlation with Multiple Symmetric Leaky Blocking Matrices for Adaptive Microphone Arrays(信号処理,LSI,及び一般)

概要

論文の詳細を見る
This paper proposes an adaptation mode controller (AMC) based on a normalized cross-correlation with multiple symmetric leaky blocking matrices (SLBMs) for adaptive microphone arrays. Robust and wideband estimation of a signal-to-interference ratio (SIR) is performed by the use of multiple pairs of SLBMs. The SIR estimate is compared to theoretical thresholds to evaluate target-interference proportion for controlling filter-coefficient adaptation in the adaptive path of the beamformer. A design method for SLBMs is developed and a design example is demonstrated. Phase differences between the outputs of the paired SLBMs are analyzed to show their desirable responses. Evaluations with recorded signals show that the proposed AMC achieves better discrimination between speech-dominant and interference-dominant sections, resulting in desirable control of coefficient adaptation in the microphone array and a better output. The better output signal of the microphone array provides a higher speech recognition rate by as much as 11% than the conventional AMC. The increased computations are shown to be negligeble.
社団法人電子情報通信学会の論文
2006-06-16