Metric Learning with Background Noise Class for Few-Shot Detection of Rare Sound Events

Abstract

Few-shot learning systems for sound event recognition gain interests since they require only a few examples to adapt to new target classes without fine-tuning. However, such systems have only been applied to chunks of sounds for classification or verification. In this paper, we aim to achieve few-shot detection of rare sound events, from long query sequence that contain not only the target events but also the other events and background noise. Therefore, it is required to prevent false positive reactions to both the other events and background noise. We propose metric learning with background noise class for the few-shot detection. The contribution is to present the explicit inclusion of background noise as a independent class, a suitable loss function that emphasizes this additional class, and a corresponding sampling strategy that assists training. It provides a feature space where the event classes and the background noise class are sufficiently separated. Evaluations on few-shot detection tasks, using DCASE 2017 task2 and ESC-50, show that our proposed method outperforms metric learning without considering the background noise class. The few-shot detection performance is also comparable to that of the DCASE 2017 task2 baseline system, which requires huge amount of annotated audio data.

Few-Shot Emergency Siren Detection

“…Five different few-shot learning methods [21,[31][32][33][34], improved with an attentional similarity module to detect transient events, are applied to sound event recognition in [35]. The effectiveness of few-shot techniques in sound event detection has led to the development of strategies to extend their application to increasingly challenging tasks, such as multi-label classification [36], rare sound event detection [37], continual learning [38], unsupervised and semi-supervised learning approaches [39], and sound localization [40].…”

Learning from Very Few Samples: A Survey

“…information extraction [211], matchine translation [100], charge prediction [178], sequence labeling [349] Audio&Speech audio/speech/sound classification [350], [351], [352], [353], [354], [355], text-to-speech [356], [357], [358], [359], acoustic/sound event detection [360], [361], [362], speech generation [350], [363], keyword/command recognition [364], keyword spotting [365], human-fall detection [366], speaker recognition [367],…”

Unsupervised Contrastive Learning of Sound Event Representations

“…Recently, [11] proposes two pretext tasks, namely, estimating the time distance between pairs of audio segments, and reconstructing a spectrogram patch from past and future patches. In the last few years, self-supervised learning methods using contrastive losses have gained increasing attention, not only for images [9,10], but also for speech [13,14,15], and sound events [16]. In the context of contrastive learning, a recent trend is to learn representations by contrasting different versions or views of the same data example, computed via data augmentation.…”