Speakers Localization Using Batch EM In Unfolding Neural Network

TL;DR

This paper introduces an interpretable neural network that integrates the Batch-EM algorithm for improved speaker localization accuracy and robustness in reverberant environments.

Contribution

It presents a novel unfolded neural network architecture embedding Batch-EM, enhancing robustness and reducing initialization sensitivity in speaker localization.

Findings

Outperforms classical Batch-EM in accuracy

Demonstrates robustness in reverberant conditions

Provides an interpretable neural network model

Abstract

We propose an interpretable Batch-EM Unfolded Network for robust speaker localization. By embedding the iterative EM procedure within an encoder-EM-decoder architecture, the method mitigates initialization sensitivity and improves convergence. Experiments show superior accuracy and robustness over the classical Batch-EM in reverberant conditions.