Contrastive Learning With Audio Discrimination For Customizable Keyword Spotting In Continuous Speech

Abstract

Customizable keyword spotting (KWS) in continuous speech has attracted increasing attention due to its real-world application potential. While contrastive learning (CL) has been widely used to extract keyword representations, previous CL approaches all operate on pre-segmented isolated words and employ only audio-text representations matching strategy. However, for KWS in continuous speech, co-articulation and streaming word segmentation can easily yield similar audio patterns for different texts, which may consequently trigger false alarms. To address this issue, we propose a novel CL with Audio Discrimination (CLAD) approach to learning keyword representation with both audio-text matching and audio-audio discrimination ability. Here, an InfoNCE loss considering both audio-audio and audio-text CL data pairs is employed for each sliding window during training. Evaluations on the open-source LibriPhrase dataset show that the use of sliding-window level InfoNCE loss yields comparable per

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyxi2024contrastive

Related papers

• Phoneme-level Contrastive Learning For User-defined Keyword Spotting With Flexible Enrollment (2024)6.34
• Sequence Discriminative Training For Deep Learning Based Acoustic Keyword Spotting (2018)8.35
• Contrastive Augmentation: An Unsupervised Learning Approach For Keyword Spotting In Speech Technology (2024)9.92
• Llm-synth4kws: Scalable Automatic Generation And Synthesis Of Confusable Data For Custom Keyword Spotting (2025)2.26
• DCCRN-KWS: An Audio Bias Based Model For Noise Robust Small-footprint Keyword Spotting (2023)5.24
• Exploring Representation Learning For Small-footprint Keyword Spotting (2023)3.58
• Slick: Exploiting Subsequences For Length-constrained Keyword Spotting (2024)5.24
• Streaming Keyword Spotting Boosted By Cross-layer Discrimination Consistency (2024)6.34