Offline High-speed Continuous Digital Speech Recognition System based on WFST

In this paper, aiming at the scenario of soldiers' fast continuous digital speaking in the field of military communication, the acoustic model and language model for fast continuous digital speech recognition are constructed based on the framework of weighted finite state machine decoder design system, which is the mainstream in the field of speech recognition at present. Then, more than 2000 hours of corpus are collected by special collector. Finally, kaldi open source development tool is used to design and implement an off-line high-speed continuous digital speech recognition system with word accuracy of more than 98%.