Offline Thai speech recognition framework on mobile device

Abstract

In this paper, we presented the offline speech recognition framework on a mobile device. The energy-based speech/silence detection is also implemented to reduce the computational workload and time. We demonstrate the performance in term of computational capability and recognition accuracy on the mobile device. The results show that the proposed offline system achieve the lower RTF by 24% compared with our previous online system on the mobile device. Furthermore, the application's startup time can reduce by using n-gram LM. In term of recognition performance, it is seen that there are no opposing effects of real environment with our proposed offline speech recognition framework.