Research On Continuous Speech Command Recognition Technology Based On Aerocraft

The technology of speech command recognition based on aerocraft in the aerial domain is studed in the paper. The purpose is to enrich the control mode of aerocraft, to lighten the pilot handling intensity, to improve manipulative speed and security. The paper has important practical significance and value to development of our country aviation project.Firstly, the reasons why the performance of the traditional speech endpoint detection algorithm which is applied to the aerial background is poor is analyzed in the paper. Then, a speech boundary detection algorithm with self-adjusting threshold is proposed based on the studies of the aerial noise from the view of intensity and type of the noise. The simulation results indicate that the performance of this endpoint decetion algorithm surpasses the traditional algorithm obviously under every kinds of noise environment. Moreover, the computation of traditional feature extract algorithm is so big as to affect system real-time performance. The algorithm is improved for the problem and parameters are optiumized according to many experiments. The computation is reduced by 50% with only 0.5% decline in recognition rate, which has almost no effect to application when the improved algorithm is adopted.The application of hidden Markov model in speech recognition is introduced, the principle and realization of the acoustics model, language model and searching algorithm are disscussed. To overcome the shortcoming of traditional Viterbi beam searching algorithm that uses fixed pruning threshold, the self-adjusting pruning threshold searching algorithm is applied in the paper and the resultant anylysis indicate that this algorithm is preferred.Finally, a small continuous speech command recognition system based on aerocraft with Visual C++ is build in the paper, of which the parameters are optimized and the average recognition rate achieves 98.5%.