Text Anchor Based Metric Learning for Small-footprint Keyword Spotting

TL;DR

This paper introduces a text anchor based metric learning approach and a novel LG-Net model to improve small-footprint keyword spotting accuracy, stability, and long-term feature modeling, achieving state-of-the-art results.

Contribution

It proposes using text anchors for more stable metric learning and designs LG-Net to better capture long-term acoustic features in KWS.

Findings

Text anchor based metric learning outperforms speech anchor methods.

LG-Net achieves SOTA accuracy of 97.67% and 96.79% on two datasets.

Lightweight LG-Net with 74k parameters attains high accuracy.

Abstract

Keyword Spotting (KWS) remains challenging to achieve the trade-off between small footprint and high accuracy. Recently proposed metric learning approaches improved the generalizability of models for the KWS task, and 1D-CNN based KWS models have achieved the state-of-the-arts (SOTA) in terms of model size. However, for metric learning, due to data limitations, the speech anchor is highly susceptible to the acoustic environment and speakers. Also, we note that the 1D-CNN models have limited capability to capture long-term temporal acoustic features. To address the above problems, we propose to utilize text anchors to improve the stability of anchors. Furthermore, a new type of model (LG-Net) is exquisitely designed to promote long-short term acoustic feature modeling based on 1D-CNN and self-attention. Experiments are conducted on Google Speech Commands Dataset version 1 (GSCDv1) and 2 (GSCDv2). The results demonstrate that the proposed text anchor based metric learning method shows consistent improvements over speech anchor on representative CNN-based models. Moreover, our LG-Net model achieves SOTA accuracy of 97.67% and 96.79% on two datasets, respectively. It is encouraged to see that our lighter LG-Net with only 74k parameters obtains 96.82% KWS accuracy on the GSCDv1 and 95.77% KWS accuracy on the GSCDv2.