Local Density-Based Anomaly Score Normalization for Domain Generalization

2025-09-13

AI Summary

Under domain shift, anomaly sound detection (ASD) suffers from distribution mismatch of anomaly scores between source and target domains—caused by acoustic environment disparities and imbalanced training data sizes—leading to poor generalization of fixed decision thresholds. To address this, we propose a local-density-based adaptive normalization method for anomaly scoring: it jointly leverages distance-based outlier detection and local density estimation in the embedding space to calibrate anomaly scores across domains. Our approach requires neither target-domain labels nor model fine-tuning, significantly enhancing the robustness and cross-domain generalization of a single model on multiple unseen acoustic environments. Extensive experiments on standard ASD benchmarks demonstrate that our method consistently outperforms existing normalization strategies, with particularly substantial gains under severe domain shifts.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Search and Optimization: Local Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

State-of-the-art anomalous sound detection (ASD) systems in domain-shifted conditions rely on projecting audio signals into an embedding space and using distance-based outlier detection to compute anomaly scores. One of the major difficulties to overcome is the so-called domain mismatch between the anomaly score distributions of a source domain and a target domain that differ acoustically and in terms of the amount of training data provided. A decision threshold that is optimal for one domain may be highly sub-optimal for the other domain and vice versa. This significantly degrades the performance when only using a single decision threshold, as is required when generalizing to multiple data domains that are possibly unseen during training while still using the same trained ASD system as in the source domain. To reduce this mismatch between the domains, we propose a simple local-density-based anomaly score normalization scheme. In experiments conducted on several ASD datasets, we show that the proposed normalization scheme consistently improves performance for various types of embedding-based ASD systems and yields better results than existing anomaly score normalization approaches.

Problem

• Normalizing anomaly scores across domains
• Addressing domain mismatch in sound detection
• Improving generalization with local density normalization

Innovation

• Local density-based normalization for scores
• Reduces domain mismatch in distributions
• Improves performance across multiple datasets