TR-SLT-0016 :2002.08.14

Kaisheng Yao, Kuldip K. Paliwal, and Satoshi Nakamura

Robust speech recognition: noise adaptive speech recognition, Sequential Monte Carlo and generative factor analyzed HMM

Abstract:This report presents works on robust speech recognition carried out by the authors. The reports have two major works. The first is noisy speech recognition in non-stationary environments, which include noise adaptive speech recognition based on sequential Kullback proximal algorithm, and sequential noise compensation based on sequential Monte Carlo method. The second part is on acoustic modeling, with a work denoted as generative factor analyzed HMM.