摘要
麦克风阵列语音增强技术已在语音信号处理中得到广泛研究.但在智能家居等领域中,当说话人等声源移动导致声源方向相对麦克风阵列发生变化时,由此产生的信号泄露将严重影响消噪性能.本文提出了一种基于麦克风阵列进行移动声源方向跟踪和噪声消除的联合算法,该方法利用两个平行滤波器组构成梯度迭代的移动声源方向跟踪器,并将该声源跟踪器嵌入广义旁瓣抵消器结构中从而进行声源方向梯度迭代跟踪和麦克风阵列语音增强的联合实现.在实验室噪声背景下移动声源的语音增强结果表明了该方法的有效性.
Microphone array speech enhancement methods such as generalized sidelobe canceller （GSC） algorithm GSC have been widely investigated and applied in fields of audio signal processing. However, when speakers move, the performance of the classic GSC algorithms will seriously deteriorate by leaking of the speech due to errors of the direction of speaker. In this paper, a novel method, which uses two parallel filters embedded in GSC structure to track the direction of the speaker,is proposed to ensure the de- noise performance of the GSC algorithm at the presence of moving speaker. The direction tracking of moving speakers is iteratively updated by gradient descent to facilitate the low complexity implementation. Experimental speech-enhancement results under moving speakers are provided to validate the effectiveness of the proposed method.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2015年第4期551-555,共5页
Journal of Xiamen University：Natural Science
基金
国家自然科学基金(61105026)
福建省高校产学合作重大项目(2015H6019)
关键词
语音增强
麦克风阵列
梯度迭代
广义旁瓣抵消器
speech enhancement
microphone array
gradient iteration
generalized sidelobe canceller
作者简介
通信作者：ftong@xmu．edu．cn
