基于多带谱相减的语音端点检测算法  

Speech Endpoint Detection Method Based on Multi-band Spectral Subtraction

作　　者：李圆[1] 赵振东[1] 杨超[1] 

机构地区：[1]华北电力大学电子与通信工程学院,河北保定071003

出　　处：《通信技术》2007年第11期353-355,共3页Communications Technology

摘　　要：为了提高噪声环境下语音端点检测的鲁棒性,介绍了一种基于多带谱相减的语音端点检测算法。仿真结果表明,与传统的基于短时能量,过零率语音端点检测算法相比,在低信噪比环境下,该算法仍可以有效的检测出起止端点。In order to detect more robustly and precisely the endpoints of speech under noisy environments, a novel endpoints detection method is proposed based on multi-band spectrum subtraction theory. Simulation results show that, as compared with the traditional algorithm based on the short-time energy or zero crossing rate, the proposed method can effectively detect the endpoints of speech at low SNR.

关 键 词：多带谱相减 端点检测 短时能量 过零率 

分 类 号：TN912.3[电子电信—通信与信息系统]