摘要
该文介绍了我们基于HMM模型语音信号识别原理开发的连续数字音识别软件包及其程序实现。其中在分割连续数字音时使用了倒谱距离法 ,取得了比传统的过零率、能量谱法更好的分割效果。最后通过一电话号码的识别demo简单给出了该连续数字音识别工具箱的应用实例。
Based on the HMM-based continuos number recognition, a software package is developed. During the process of segmenting, a better segmentation has been obtained by adopting the endpoint detection method of cepsrtal distance .In order to demonstrate the implement of the functions in the software package, a telephone number recognition example is given.
出处
《计算机仿真》
CSCD
2002年第5期81-83,共3页
Computer Simulation
基金
甘肃省自然科学基金 2S0 11-A2 5 -0 16 -G
