Development of Articulation Training System with Speech Recognition Based Automatic Pronunciation Detection Mechanism

The deficiency of expressing symbol ability of the spoken language for articulation disorder will leads to the difficulty of verbal expression and communication. In the clinical protocol, language therapist subjectively utilizes the clinical experience to individually test and treat articulation disorder. The manpower of language therapist and the computer based assistant instructions are also insufficient. In this paper, an articulation training system with automatic pronunciation detection mechanism is proposed to assistance in the program of language treating for language therapist. The articulation errors in phonetic level are analyzed and modeled by clinical linguist. According to the articulation errors, a speech recognition based automatic pronunciation detection algorithm is developed to effectively acquire use’s pronunciation characteristic. Speechreading based feedback responses are designed and applied to improve the efficiency of training program. Preliminary results reveal the practicability of our proposed method and system.