摘要
本文介绍一种新的基于麦克风阵列的语音分离和说话人跟踪技术 .该技术使用麦克风阵列 ,形成一个指向感兴趣说话人的波束来增强信号 ,并通过方向置零来抑制其他说话人的声音和噪声 ,同时用自适应算法跟踪说话人的方位变化 .仿真验证了该技术的有效性 .与常规的自适应算法相比 ,该算法不需训练序列 ,具有显著的优势 .
A new speech sources separation and speaker tracking technique is introduced based on microphone arrays. By means of spatial property of the received speech signals from microphone arrays, this method utilizes beamforming to estimate the DOA of the speaker of interest, and attenuates unwanted voices by nulling other directions. Considering the speech environments where the speaker may freely move and the background voices exist, an adaptive algorithm is used to track the movements and the source direction variations automatically. Computer simulations validate the effectiveness of the technique. Compared with the conventional methods, the scheme needs no training sequence, and have great potential practical advantages.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2005年第2期382-384,共3页
Acta Electronica Sinica
关键词
麦克风阵列
语音分离
说话人跟踪
波束形成
Acoustic wave propagation
Adaptive algorithms
Computer simulation
Microphones
Neural networks
Signal processing
