Approximate MVDR and MMSE Beamformers Exploiting Scale-Invariant Reconstruction of Signals on Microphones
IWAENC(2016)
摘要
Minimum Variance Distortionless Response (MVDR) and Minimum Mean-Squared Error (MMSE) beamformers are popular array processors for enhancing multichannel recordings of a directional source. We propose their approximate variants having the generalized sidelobe canceler structure whose performances depend purely on the blocking matrix part. No auxiliary methods such as adaptive interference canceler or voice activity detector to estimate the source/noise covariance are needed. Instead, scale-invariant least square estimators are used, which enable to estimate the noise also during speech activity and to recover original spectra of the target source on the microphones. In experiments, we compare signal-to-noise ratio improvements of several variants of the beamformers achieved on six-channel recordings of speech in nonstationary noisy conditions.
更多查看译文
关键词
Minimum Variance Distortionless Beamformer,Minimum Mean Square Error Beamformer,Blind Source Separation,Scaling Ambiguity
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要