中文摘要：

在语音信号处理领域,端点检测作为语音信号处理的初始化工作,有效合理的端点检测方法不仅可以减少数据的存储量和处理时间,而且可以排除无声段的噪声干扰,使语音的后续处理更加准确. 本文为医疗语音信号处理提供一种医院噪声环境下医疗语音信号的端点检测方法. 综合采用频域下的特征倒谱距离、能量谱、能量谱方差、统计学特征谱熵值4种特征作为控制条件判断语音的端点. 通过设计软件验证了算法的有效性.

英文摘要：

In the field of speech signal processing, endpoint detection as the initialization of voice signal processing, effective and reasonable endpoint detection method can not only reduce the data storage capacity and processing time, but it can also eliminate the noise of the noise segment, so that the subsequent processing of voice can become more accurate. A method of detecting the endpoint of medical voice signal in noise environment is proposed, which provides an accurate endpoint detection method for medical voice signal processing in hospital noise environment. The endpoints of speech are judged by four characteristics including crosstalk distances, energy spectrum, energy spectrum variance and statistical characteristic spectral entropy. The effectiveness of the algorithm is verifed by the design software.