Dummy Prototypical Networks for Few-Shot Open-Set Keyword Spotting

About

Keyword spotting is the task of detecting a keyword in streaming audio. Conventional keyword spotting targets predefined keywords classification, but there is growing attention in few-shot (query-by-example) keyword spotting, e.g., N-way classification given M-shot support samples. Moreover, in real-world scenarios, there can be utterances from unexpected categories (open-set) which need to be rejected rather than classified as one of the N classes. Combining the two needs, we tackle few-shot open-set keyword spotting with a new benchmark setting, named splitGSC. We propose episode-known dummy prototypes based on metric learning to detect an open-set better and introduce a simple and powerful approach, Dummy Prototypical Networks (D-ProtoNets). Our D-ProtoNets shows clear margins compared to recent few-shot open-set recognition (FSOSR) approaches in the suggested splitGSC. We also verify our method on a standard benchmark, miniImageNet, and D-ProtoNets shows the state-of-the-art open-set detection rate in FSOSR.

Byeonggeun Kim, Seunghan Yang, Inseop Chung, Simyung Chang · 2022

Related benchmarks

Task	Dataset	Result
Few-shot Audio Classification	FSC-89, NSynth-100, and LS-100 Generalizability Cross-Dataset 5-way 5-shot	Accuracy71.19	54
Few-shot classification	FSC-89 → NSynth-100	Accuracy70.13	31
Few-shot classification	FSC-89 → LS-100	Accuracy71.19	22
Few-shot classification	FSC-89 to NSynth-100	AUROC0.5816	18
Few-shot Open-set Audio Classification	Domestic Environments	Accuracy80.62	18
Few-shot classification	FSC-89 to LS-100	AUROC63.03	9
Few-shot classification	LS-100 to NSynth-100	AUROC65.19	9
Few-shot classification	LS-100 → FSC-89	Accuracy (FSC-89 Few-shot)44.44	9
Few-shot classification	LS-100 → NSynth-100	Accuracy65.76	9
Open-set Few-shot Classification	NSynth-100 5-way 1-shot	Accuracy86.4	9