Streaming Keyword Spotting Boosted by Cross-layer Discrimination Consistency

TL;DR

This paper introduces a novel streaming decoding algorithm with cross-layer discrimination consistency for CTC-based keyword spotting, significantly improving recall and reducing false alarms in noisy and clean conditions.

Contribution

It proposes a new decoding method that detects keywords at any position and leverages cross-layer info to enhance discrimination, outperforming existing ASR and graph-based methods.

Findings

Achieves 6.8% absolute recall improvement

Reduces miss rate by 46.3% relative

Maintains a low false alarm rate of 0.05 per hour

Abstract

Connectionist Temporal Classification (CTC), a non-autoregressive training criterion, is widely used in online keyword spotting (KWS). However, existing CTC-based KWS decoding strategies either rely on Automatic Speech Recognition (ASR), which performs suboptimally due to its broad search over the acoustic space without keyword-specific optimization, or on KWS-specific decoding graphs, which are complex to implement and maintain. In this work, we propose a streaming decoding algorithm enhanced by Cross-layer Discrimination Consistency (CDC), tailored for CTC-based KWS. Specifically, we introduce a streamlined yet effective decoding algorithm capable of detecting the start of the keyword at any arbitrary position. Furthermore, we leverage discrimination consistency information across layers to better differentiate between positive and false alarm samples. Our experiments on both clean and noisy Hey Snips datasets show that the proposed streaming decoding strategy outperforms ASR-based and graph-based KWS baselines. The CDC-boosted decoding further improves performance, yielding an average absolute recall improvement of 6.8% and a 46.3% relative reduction in the miss rate compared to the graph-based KWS baseline, with a very low false alarm rate of 0.05 per hour.