低信噪比环境下子带能熵比端点检测算法
    点此下载全文
引用本文:沈钰瑞,李文钧,金伟杰,岳克强.低信噪比环境下子带能熵比端点检测算法[J].计算技术与自动化,2020,(2):109-113
摘要点击次数: 8
全文下载次数: 0
作者单位
沈钰瑞,李文钧,金伟杰,岳克强 (杭州电子科技大学 射频电路与系统教育部重点实验室浙江 杭州 310018) 
中文摘要:语音端点检测是将采集到的语音信号从复杂的噪声背景中提取出来,确定每段语音的开始和结束,是后续处理的基础。对于语音端点检测在低信噪比的复杂噪声环境下准确率低的问题,提出了一种多窗谱估计减噪和子带能熵比法结合的语音端点检测算法。该算法通过改进多窗谱谱减法对语音信号进行减噪,在分析了常规谱熵端点检测算法的基础上结合对数能量,以改进的子带能熵比作为阈值进行端点检测。实验表明,该算法在不同环境的低信噪环境下,准确率高,具有较高的鲁棒性。
中文关键词:端点检测  低信噪比  谱熵法  子带能熵比法  对数能量
 
Researoh of Energy-entropy-ratio Endpoint Detection Algorithm in Low SNR Environment
Abstract:The speech endpoint detection is to extract the speech segment from the complex noise background in the collected speech signal,and determine the start and end of each speech,which is the basis of subsequent processing. For the problem that speech endpoint detection has low accuracy under complex noise environment with low SNR,a speech endpoint detection algorithm based on multi-taper spectrum estimation and noise reduction combined with improved energy entropy ratio method is proposed. The algorithm denoises the speech signal by improving the multi-taper spectral subtraction. Based on the analysis of the conventional spectral entropy endpoint detection algorithm combined with the logarithmic energy,the improved subband entropy ratio is used as the threshold for endpoint detection. Experiments show that the algorithm has high accuracy and high robustness under low SNR environment in different environments.
keywords:endpoint detectionlow  low signal to noise ratio  spectral entropy  subband energy entropy ratio  logarithmic energy
查看全文   查看/发表评论   下载pdf阅读器