민원서류 발급용 화자독립 대용량 음성인식 시스템에 관한 연구

A Study on the Large Vocabulary Based Speech Recognition System for a Cyber Civil Affairs

On issuing automatically various civil affairs documents, we have a little problem with keyboard input system. It has the only way how to input system which consist of a button touch or screen touch. Nowadays, we need to a new paradigm of a interface technology. Also, the large vocabulary isolated word recognition system is required as the optimal HMI technique. For the proposed input terminal with a highly limited processing time and memory capacity, it is impossible to process more than 100,000 words in the input terminal. Therefore, we proposed phoneme recognizer using a phonetic GMM and an PDM Levenshtein distance with multi-layer architecture for the speech recognition of input terminal. By the proposed methods, we obtained high performance in the input terminal with low speed processing and small memory capacity. We obtained the recognition rate of maximum 94.8% in indoor environment and the maximum 92.4% in the outdoor office environments.