Nick CAMPBELL
The ATR British English Speech Database
Abstract:The ATR English Speech database consists of six sets of speech data prepared especially for recognition and synthesis research to illustrate all combinations of the speech sounds of English in a variety of phonemic and prosodic contexts. They consist of readings by four RP speakers of the 5000 most frequent words, the alphabet names, the cardinal numbers, the radio-telephony alphabet, and a set of 200 phonemically-balanced sentences. The sentences were read in two styles: as isolated words and as continuous speech. The isolated-word readings maintain some of the sentence-level intonation, and are prosodically different from the same words read as a list, but have a clear pause between each pair of words.
The readers are two male (ma, mb) and two female (fa, fb) young adults, who have been resident in the United Kingdom all their lives. They were selected from a panel of candidates by a trained phonetician and have all been judged to speak the standard dialect without disfluencies.