Tools for speech perception, production, and training studies: Web‐based second language training system, and a speech resynthesis system

We describe a Web‐based instruction system for second language (L2) learners, dubbed ATR CALL. This system is a collection of speech perception, production, and comprehension training tools, designed to examine L2 acquisition mechanisms, especially for native speakers of Japanese who are learning English as L2. Each training component focuses on the acoustic‐phonetic, prosodic, lexical, or semantic decoding level of spoken language. The stimuli are selected and presented from a large speech database of 13 000 English words and 6000 English sentences. In the pronunciation training component, learners’ productions are scored using HMM‐based evaluation algorithms, and automatic and immediate feedback about pronunciation goodness is provided. Speech analysis tools for language learners are also provided. We also describe a speech manipulation system, dubbed STRAIGHT [Kawahara et al., Speech Commun. (1999)]. This system resynthesizes stimuli with quality as high as the original stimulus even after manipulat...