摘要
根据离散余弦变换(DCT)特性和最大离散熵定理,该文提出利用短时语音余弦变换谱的谱熵进行语音信号检测(Voice activity detection)。在强噪声环境下,传统的能量,过零率,相关等检测方法将会失效,而谱熵法则具有稳健的抗噪特性。
An approach is introduced to voice detection that differs from normal approach via energy, correlation and zerocrossing criteria. By measuring the gross shape of the short-term speech spectrum using spectral entropy to detect voice segment, it is shown that the spectral entropy can be used effectively even in heavy background noise. The simulation results show that the approach via spectral entropy has good performance for anti-noice.
出处
《电子与信息学报》
EI
CSCD
北大核心
2001年第10期989-993,共5页
Journal of Electronics & Information Technology
关键词
语音检测
DCT变换
谱熵
抗噪
Voice detection, DCT transformation, Spectral entropy, Robustness
