Temporal Pooling Strategies for Training-Free Anomalous Sound Detection with Self-Supervised Audio Embeddings

2026-03-04

arXiv.org

AI Summary

This work addresses the limitations of existing training-free anomaly sound detection (ASD) methods, which predominantly rely on temporal mean pooling and overlook the potential of alternative pooling strategies. The study systematically evaluates various temporal pooling mechanisms and introduces an adaptive Relative Deviation Pooling (RDP) scheme, along with a hybrid approach combining RDP and Generalized Mean (GeM) pooling. Notably, the proposed method significantly enhances detection performance without requiring fine-tuning of pretrained audio embedding models. Extensive experiments across five benchmark datasets demonstrate that the approach substantially outperforms conventional mean pooling, achieving state-of-the-art results among training-free ASD methods—and in some cases even surpassing trained or ensemble-based approaches.

Abstract

Training-free anomalous sound detection (ASD) based on pre-trained audio embedding models has recently garnered significant attention, as it enables the detection of anomalous sounds using only normal reference data while offering improved robustness under domain shifts. However, existing embedding-based approaches almost exclusively rely on temporal mean pooling, while alternative pooling strategies have so far only been explored for spectrogram-based representations. Consequently, the role of temporal pooling in training-free ASD with pre-trained embeddings remains insufficiently understood. In this paper, we present a systematic evaluation of temporal pooling strategies across multiple state-of-the-art audio embedding models. We propose relative deviation pooling (RDP), an adaptive pooling method that emphasizes informative temporal deviations, and introduce a hybrid pooling strategy that combines RDP with generalized mean pooling. Experiments on five benchmark datasets demonstrate that the proposed methods consistently outperform mean pooling and achieve state-of-the-art performance for training-free ASD, including results that surpass all previously reported trained systems and ensembles on the DCASE2025 ASD dataset.

Problem

• anomalous sound detection
• training-free
• temporal pooling
• self-supervised audio embeddings
• pre-trained models

Innovation

• temporal pooling
• training-free anomalous sound detection
• self-supervised audio embeddings
• relative deviation pooling
• generalized mean pooling