Independent Vector Extraction for Fast Joint Blind Source Separation and Dereverberation

We address a blind source separation (BSS) problem in noisy reverberant environment which the number of microphones $M$ is greater than sources interest, and other noise components can be approximated as stationary Gaussian distributed. Conventional BSS algorithms for optimization multi-input multi-output convolutional beamformer have suffered from huge computational cost when large. here propose computationally efficient method that integrates weighted prediction error (WPE) dereverberation fast called independent vector extraction (IVE), has been developed less environments. show that, given power spectrum each source, new reduced to IVE by exploiting condition, makes easy handle efficient. An experiment speech signal shows compared conventional WPE analysis, our proposed achieves much faster convergence while maintaining its performance.