Multichannel Keyword Spotting for Noisy Conditions

TL;DR

This paper introduces a neural network-based multichannel keyword spotting method with attention mechanisms, enhancing performance in noisy environments by effectively selecting and combining input channels, validated on controlled and natural datasets.

Contribution

It proposes a novel neural network architecture with attention for multichannel KWS, improving noise robustness without degrading signal quality, outperforming existing solutions.

Findings

Enhanced noise reduction metrics compared to baselines

Improved keyword spotting accuracy in noisy conditions

Efficient use of computational resources

Abstract

This article presents a method for improving a keyword spotter (KWS) algorithm in noisy environments. Although beamforming (BF) and adaptive noise cancellation (ANC) techniques are robust in some conditions, they may degrade the performance of the activation system by distorting or suppressing useful signals. The authors propose a neural network architecture that uses several input channels and an attention mechanism that allows the network to determine the most useful channel or their combination. The improved quality of the algorithm was demonstrated on two datasets: from a laboratory with controlled conditions and from smart speakers in natural conditions. The proposed algorithm was compared against several baselines in terms of the quality of noise reduction metrics, KWS metrics, and computing resources in comparison with existing solutions.