Temporal Pooling Strategies for Training-Free Anomalous Sound Detection with Self-Supervised Audio Embeddings
For researchers in anomalous sound detection, this work provides a simple yet effective improvement to training-free ASD methods, though it is an incremental advance over existing embedding-based approaches.
The paper systematically evaluates temporal pooling strategies for training-free anomalous sound detection (ASD) using pre-trained audio embeddings, proposing relative deviation pooling (RDP) and a hybrid strategy that consistently outperform mean pooling and achieve state-of-the-art results, surpassing previously reported trained systems on the DCASE2025 ASD dataset.
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 assigns larger weights to embeddings with stronger temporal deviations, and introduce a hybrid pooling strategy that combines RDP with generalized mean (GeM) 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 previously reported trained systems and ensembles on the DCASE2025 ASD dataset.