摘要
本文应用对数面积比参数,进行了非限定内容话者辨识的实验研究。应用语噪声法获得了稳定的个人语音长时间平均谱。实验用语音为任意找来的10s左右的一小段文章。16名男性、4名女性参加了实验,平均识别率为80%,且话者落在前5名候选人中的概率为95%以上。
The experiment of text-independent speaker recognition is used by
the Log Area Ratio and the steady longtime average speech spectctrum
of personal speech is obtained by speech noise in this paper. the mater-
ials used in the experiment are arbitrory. And the selected small para-
graph of articles lasts about 10 minutes. 16 males and 4 females took
park in the experiment, the average recognition on rate is 80%, and the
probability that the recognition speaker may be in the superior 5 candi-
dates is above 95%.
关键词
语音识别
对数面积比
语噪声法
phonetic recognition
logarithm area ratio
speech noise method