Research on the Chinese initial pronunciation computer-aided language learning system

In order to facilitate the language rehabilitation of the hearing impaired person,an automatic speech recognition-based computer-aided language learning(CALL)system is designed.Firstly,an utterance database produced by a standard spoken teacher is established and the key pronunciation mouth shape frames are recorded.Secondly,the static and dynamic features are extracted from the utterance database and we apply the continuous density hidden markov model(CDHMM) to imitate Chinese initial,which is trained with embedded training algorithm and performs recognition task with token passing algorithm and recognition network.The outside test recognition rate of the teacher pronunciation is 96.65%,which means that the system can reasonably work as the benchmark system of the Chinese initial pronunciation learning.Finally,the performance comparison experiment based on visual feedback proves that the system can aid the deaf-dumb to recover language ability effectively.