论文部分内容阅读
浊音声母的特点是低频信号显著:使用信号经过高频加重滤波器之后和之前的能量比可以反映语音信号能量集中的频率范围。据此,本文提出了一个将浊音声母从可能含有它的孤立字音中分隔出来并且进行分类和辨认的算法。对来自不同省分的八个讲者的语音数据进行分类和辨认实验,区分浊音声母和零声母两类的正确分类率达到98％以上。使用线性判别技术辨认各个浊音声母,平均正确辨认率达到86％。
The voiced consonant is characterized by a significant low-frequency signal: the energy ratio after and after the high-frequency emphasis filter is applied to the signal can reflect the frequency range over which the speech signal energy is concentrated. Accordingly, this paper presents an algorithm that separates and classifies and identifies the voiced consonants from the isolated phonograms that may contain them. The speech data of eight speakers from different provinces were classified and identified experimentally, and the correct classification rate of both phonetic and zero initials was over 98%. The use of linear discriminant techniques to identify each voiced consonant, the average correct recognition rate of 86%.