Improvements of Discriminative Feature Space Training for Anomalous Sound Detection in Unlabeled Conditions

TL;DR

This paper enhances discriminative feature space training for anomalous sound detection in unlabeled conditions by improving feature extraction and introducing pseudo-labeling methods, leading to better performance when labels are missing.

Contribution

The paper introduces a novel multi-resolution spectrogram-based feature extractor and various pseudo-labeling techniques to improve anomalous sound detection without labeled data.

Findings

Significant performance improvements with the proposed methods.

Effective feature extractor under unlabeled conditions.

Pseudo-labeling enhances training accuracy.

Abstract

In anomalous sound detection, the discriminative method has demonstrated superior performance. This approach constructs a discriminative feature space through the classification of the meta-information labels for normal sounds. This feature space reflects the differences in machine sounds and effectively captures anomalous sounds. However, its performance significantly degrades when the meta-information labels are missing. In this paper, we improve the performance of a discriminative method under unlabeled conditions by two approaches. First, we enhance the feature extractor to perform better under unlabeled conditions. Our enhanced feature extractor utilizes multi-resolution spectrograms with a new training strategy. Second, we propose various pseudo-labeling methods to effectively train the feature extractor. The experimental evaluations show that the proposed feature extractor and pseudo-labeling methods significantly improve performance under unlabeled conditions.