CH_Broadcast_ASR is a 127-hour Chinese speech recognition dataset from FM radio and TV programs, offering high-quality, fast-paced broadcast speech for training and evaluating end-to-end ASR models. - View it on GitHub