多位朝鲜语数字语音识别方法的研究

针对朝鲜语数字串语音识别中单位符号容易被误判为数字的问题，提出了基于朝鲜语数字串结构特点的识别方法．首先，对事先采集好的语音样本提取特征并存储成数字模板和单位符模板；其次，将数字串测试样本切分成孤立的数字或单位符并提取特征参数；最后，根据朝鲜语数字串发音特点，即若前1个孤立词已被识别为数字，那么后1个被默认为单位符并且只与存储单位符号的模板进行匹配．实验结果表明，该方法不仅可以获得较高的正确识别率，同时提高了识别效率．

Recognition of Multi-digit Korean Number Speech

According to the problem that Korean numbers speech recognition unit symbol is mistaken for a number, a recognition method based on characteristics of Korean digital string is proposed. First of all, the feature of speech samples were extracted and stored into a digital template and digits template; Second, cutting the string of numbers into an isolated symbol and extracting feature parameters; Finally, according to the characteristics of Korean pronunciation of the number string, that is, if the previous word has been identified as an isolated number, the unit is a default symbol for the next one and matching with only the unit template. Experimental results show that the method can not only obtain a higher recognition rate, but also improve the recognition efficiency.