Efficient Independent Vector Extraction of Dominant Target Speech

TL;DR

This paper introduces a computationally efficient blind speech extraction method that isolates a dominant target speaker using a modified independent vector analysis algorithm, suitable when the target's average power exceeds interference.

Contribution

The paper presents a novel, simplified blind speech extraction technique based on a modified independent vector analysis, addressing computational complexity and scaling ambiguity.

Findings

Effective extraction of dominant speech demonstrated in simulations.

Reduced computational complexity compared to full blind source separation.

Successful handling of scaling ambiguity with a new one-unit scaling operation.

Abstract

The complete decomposition performed by blind source separation is computationally demanding and superfluous when only the speech of one specific target speaker is desired. In this paper, we propose a computationally efficient blind speech extraction method based on a proper modification of the commonly utilized independent vector analysis algorithm, under the mild assumption that the average power of signal of interest outweighs interfering speech sources. Considering that the minimum distortion principle cannot be implemented since the full demixing matrix is not available, we also design a one-unit scaling operation to solve the scaling ambiguity. Simulations validate the efficacy of the proposed method in extracting the dominant speech.