Speech Scenario Adaptation and Discourse Topic Recognition on Mobile Smart Terminal

Context identifying based on speech data is important to social services and city management. In a complex application environment, a speech recognition system needs to address two main problems: background noises and large vocabulary search latency. We use the adjustment acoustic model to deal with the scenario adaptation, and we use adjustment dictionary and language module to solve the discourse topic recognition. As a case study, we design and implement a continuous language speech recognition system on a mobile smart terminal. Experiments show that the scene adaptation effectively improves the accuracy rate of speech recognition, and the discourse topic recognition verifies the recognition effectiveness of our speech recognition system.