基于隱馬爾可夫模型的語(yǔ)音識(shí)別技術(shù)在載人航天器上的應(yīng)用

Spacecraft speech recognition based on Hidden Markov Model (HMM) technology

摘要: 利用嵌入式平臺(tái)構(gòu)建了基于隱馬爾可夫模型（Hidden Markov Model，HMM）技術(shù)的載人航天器語(yǔ)音識(shí)別系統(tǒng)，并針對(duì)載人航天器特殊的噪聲背景和聲學(xué)訓(xùn)練，對(duì)傳統(tǒng)語(yǔ)音端點(diǎn)檢測(cè)算法進(jìn)行了改進(jìn)。通過(guò)試驗(yàn)在嵌入式運(yùn)行環(huán)境下實(shí)現(xiàn)了指令發(fā)送、文字輸入等功能，證實(shí)該系統(tǒng)適應(yīng)航天員的口音和發(fā)音習(xí)慣，可對(duì)連貫自然發(fā)音方式的命令進(jìn)行識(shí)別，能夠滿足航天員及任務(wù)使用需求。

Abstract: This paper reviews the theory of the Hidden Markov Model(HMM)and the speech recogIlition system. By using the embedded platform, the manned-spacecrar speech recognition system is built based on the HMM technology. To meet complex requirements of future spacecraft for speech recognition, the specifications and rules for the speech recognition are optimized, with an improvement of the accuracy rate over the traditional voice activity detection algorithm. The application of this continous speech recognition system on manned-spacecraft shows good results for the fhnction implementation of command sending and character input.