关键词检测系统中废料模型技术的研究

关键词检测是语音识别中一个重要的研究方向．关键词检测技术的重点之一就是代表非关键词语音的模型，不论识别系统的方法怎样，大部分是采用废料模型的，而废料模型的结构和类型对整个系统的性能有很大的影响．提出了一种基于音节格的废料模型．实验表明，与传统的基于音素类的废料模型相比，关键词的检测率有了很大的提高．

Study of garbage model technology in keyword spotting system

Keyword spotting is one of important topics in speech recognition. For the keyword spotting technology, the emphasis among them is the model which represents the non-keyword. Whatever the methods are about recognition system, most of them employ the garbage model. However, the structure and type of garbage models have an important effect on the overall system performance. The garbage model based on a syllable lattice is proposed. Compared with the traditional garbage model based on phoneme class, experimental results show that the garbage model based on a syllable lattice has a great improvement in the detection probability for the keywords.