摘要
针对在低信噪比条件下难以实现语音端点检测,提出了基于混沌理论的解决方法,采用Duffing方程的间歇混沌特性对语音信号进行检测,同时对谱减法作了改进,根据人耳听觉掩蔽效应的语音增强算法,动态修正谱减系数,有针对性地进行谱减,有效克服了音乐噪声。在信噪比较低的情况下,按照该方案处理后的语音不仅信噪比显著提高,主观听觉失真有效减小,且语音清晰度、可懂度和舒适度极大改善。
A solution based on chaos theory is proposed to overcome the difficulty of low SNR speech endpoint detection. The intermittent chaos characteristic of Duffing equation is employed to detect audio signal, and the spectral subtraction is improved. Based on masking properties of human auditory system, the voice enhancement algorithm performs selective spectral subtraction, and overcomes the music noise effectively. The experimental results show that the proposed algorithm is superior to the general spectral subtraction methods, not only in improvement of SNR and reduction of perceptual distortions, but also in voice clarity, intelligibility, and comfort.
出处
《电声技术》
2008年第1期63-66,69,共5页
Audio Engineering
关键词
信噪比
语音端点检测
混沌
听觉掩蔽效应
语音增强
SNR
voice activity detection
chaos
masking effect of human auditory system
speech enchancement
作者简介
章勇勤，硕士研究生，主要研究方向为多媒体信息处理、电子设计自动化（EDA）。
