论文部分内容阅读
本文提出了基于 3维空间Viterbi算法的汉语连续语音识别方法 .本方法采用 6 0个音素单位的隐马尔可夫模型 (HMM)和 8个声调单位的HMM作为识别用基元模型 .音素基元模型和声调基元模型的识别结果的统合 ,采用音素单位的HMM状态、声调单位的HMM状态和时间的 3维空间Viterbi算法来实现 .语音声学处理和语音言语处理的结合 ,采用修改型Earley分析法的Top Done型文法分析器和OnePassDP为基础的帧同步识别算法来实现 .在由 10名话者发音的有关旅馆预约指南的识别困难度是 2 7 3的 10 70句子的识别实验中 ,总平均识别率达到 94 4% .
In this paper, a Chinese continuous speech recognition method based on the Viterbi algorithm in 3-D space is proposed. This method uses HMM (HMM) of 60 phonemes and HMM of 8 tones as the recognition primitive model. The integration of the model and the recognition result of the tone primitive model is realized by using the HMM state of the phoneme unit, the HMM state of the tone unit and the 3-dimensional Viterbi algorithm of time. The combination of the speech acoustics and the speech speech processing is implemented by a modified Earley analysis Method of Top Done grammar analyzer and OnePass DP based frame synchronization recognition algorithm to achieve.In the 10 speaker interpretation of the hotel reservation guide recognition difficulty is 2 7 3 10 70 sentence recognition experiments, The average recognition rate reached 94 4%.