摘要
为了更好地运用深度置信网络进行语音端点检测,针对现有方法过于繁杂的问题,改进采用语音频谱作为深度置信网络的输入。在Matlab环境下使用TIMIT语料库进行仿真实验,实验结果验证了该方法的有效性,并且在Babble噪声环境下验证该方法比现有方法具有更高的检测正确率。
Focusing on existing methods are too complicated and in order to make better use of the deep belief networks for voice activity detection, an improved voice spectrum as deep belief networks’input is used in this paper. Simulation experiments use TIMIT corpus with the Matlab, experimental results verify the validity of the method, and show that this method has higher detection accuracy percentage of the examination than existing methods in Babble environments.
出处
《计算机工程与应用》
CSCD
2014年第20期207-210,共4页
Computer Engineering and Applications
基金
湖南省科技计划项目(No.2012FJ3025)
关键词
语音端点检测
深度置信网络
快速傅里叶变换
语音信号处理
voice activity detection
deep belief networks
fast Fourier transform
speech signal processing
