A comparative study of fMPE and RDLT approaches to LVCSR
ISCSLP（2012）
摘要
This paper presents a comparative study of two discriminatively trained feature transform approaches, namely feature-space minimum phone error (fMPE) and region-dependent linear transform (RDLT), to large vocabulary continuous speech recognition (LVCSR). Experiments are performed on an LVCSR task of conversational telephone speech transcription using about 2,000 hours training data. Starting from a maximum likelihood (ML) trained GMM-HMM based baseline system, recognition accuracy and run-time efficiency of different variants of the above two methods are evaluated, and a specific RDLT approach is identified and recommended for deployment in LVCSR applications.
更多查看译文
关键词
recognition accuracy,feature-space minimum phone error,speech processing,lvcsr,rdlt,speech recognition,large vocabulary continuous speech recognition,fmpe,run-time efficiency,learning (artificial intelligence),maximum likelihood estimation,discriminatively trained feature transform approaches,lvcsr task,conversational telephone speech transcription,feature transform,specific rdlt approach,discriminative training,gaussian mixture models,transforms,region-dependent linear transform,maximum likelihood trained gmm-hmm based baseline system,hidden markov models,learning artificial intelligence
AI 理解论文
溯源树
样例
生成溯源树，研究论文发展脉络
Chat Paper
正在生成论文摘要