Simulating realistic speech overlaps improves multi-talker ASR.
Muqiao YangNaoyuki KandaXiaofei WangJian WuSunit SivasankaranZhuo ChenJinyu LiTakuya YoshiokaPublished in: CoRR (2022)
Keyphrases
- automatic speech recognition
- speech recognition
- speech signal
- word error rate
- spontaneous speech
- broadcast news
- noisy environments
- hidden markov models
- conversational speech
- spoken words
- real life
- language model
- recognition engine
- emotion recognition
- endpoint detection
- recognition errors
- human machine interaction
- pattern recognition
- natural language