Using the proposed method, a 333 % reduction in word error rate is obtained relative to conventional HMM system for a speaker-independent, small vocabulary recognition task. 使用所提出的方法对不特定话者小词表进行实验,其系统误识率与基本HMM系统相比下降了333%。
The experiment on speaker independent large vocabulary continuous speech recognition is introduced, the word error rate is reduced by 6 %, which shows that the duration information improves the performance of the system. 介绍了非特定人大词汇量连续语音识别的实验,实验结果表明,利用段长信息改进识别算法比原识别算法字的误识率降低了6%。
Experimental results on a large vocabulary continues speech recognition task of mandarin show that in comparison to the traditional diagonal modeling technique, the proposed method can get nearly 18.8 % word error rate reduction without incurring much computation load during decoding. 试验结果表明,在汉语大词汇量连续语音识别系统中,同传统的对角方差建模技术相比,这种方法在计算量增加很小的情况下,系统字的误识率降低了18.8%。
Tests show that the method is able to reduce the word error rate by about 10 % even with very short utterance. 实验表明,该方法可以利用很少的自适应数据使识别的字错误率下降10%左右。
Improved Word Error Rate(WER) evaluation algorithm for automatic speech recognition 一种改进的语音识别词错误率评估算法