基于谱减法和短时零熵法的语音端点检测  被引量:2

Research of Endpoint Detection Based on Spectral Subtraction and Zero Spectral Entropy Method

在线阅读下载全文

作  者:吴鹏 章小兵 丁武 WU Peng;ZHANG Xiaobing;DING Wu(Anhui University of Technology,Ma Anshan 243000,Chin)

机构地区:[1]安徽工业大学电气与信息工程学院,安徽马鞍山24300

出  处:《电声技术》2018年第3期55-59,共5页Audio Engineering

基  金:安徽工业大学产学研基金资助重大项目(RD14206003)

摘  要:针对低信噪比环境下基于能零比的端点检测方法准确率低的问题,提出了一种结合谱减法和短时零熵法的端点检测方法。首先通过功率谱减法对带噪语音进行降噪处理,然后结合短时平均过零率特征与谱熵特征,构造出一种新的语音参数,即短时零熵,最后运用双门限算法进行端点检测。实验结果表明该方法在复杂噪声环境下具有良好的端点检测性能,降低了检测的误判率。In order to solve the problem of low accuracy of endpoint detection based on energy-to-zero ratio in low SNR environment,an endpoint detection method is put forward,combining spectral subtraction and short-time zero entropy method. Firstly,the power spectrum subtraction is used to denoise the noisy speech,and then a new speech parameter is constructed by combining the characteristics of the short-term average zero-crossing rate and the spectral entropy,which is named Short-term Zero Spectral Entropy. Finally,double threshold algorithm is used to achieve endpoint detection. The simmulation results show that this method has good endpoint detection performance under the environment of complex background noise and reduces the false positive rate of detection.

关 键 词:短时零熵 谱减法 双门限算法 端点检测 

分 类 号:TN912.3[电子电信—通信与信息系统]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象