摘要
端点检测和声韵母分离是汉语语音信号处理的一个重要预处理步骤。针对单门限算法在检测过程中出现的漏检问题,提出了一种基于双门限算法的汉语语音信号端点检测和单音节声韵母检测方法。通过短时能量与过零率相结合,弥补单门限算法的不足,并对检测方法进行仿真,验证了该方法的准确性和有效性。
Endpoint detection and initial/final segmentation is a important preprocessing step of Chinese speech signal processing.Aiming at the omission occurred in the endpoint detection based on single-threshold algorithm,a new method based on the dual-threshold algorithm of Chinese speech signal endpoint detection and the monosyllable initial/final segmentation was presented.The new method combined short-time energy with zero-crossing rate for making up for the shortcoming of single-threshold algorithm.The simulation experiments for the detection method prove that the method is accurate and effective.
出处
《桂林电子科技大学学报》
2011年第6期480-484,共5页
Journal of Guilin University of Electronic Technology
基金
国家自然科学基金(60961002)
广西教育厅科研基金(201010LX454)
关键词
端点检测
声韵母分离
双门限算法
仿真
endpoint detection
initial/final segmentation
dual-threshold algorithm
simulate
