Anomalous sound detection based on interpolation deep neural network

TL;DR

This paper introduces an interpolation-based deep neural network method for anomalous sound detection that improves accuracy, especially for non-stationary machinery sounds, by predicting missing spectrogram frames.

Contribution

It proposes a novel interpolation approach using multiple spectrogram frames to enhance anomaly detection in non-stationary sounds, outperforming traditional autoencoder methods.

Findings

Achieved 27% improvement in AUC score over conventional methods.

Effective in detecting anomalies in non-stationary machinery sounds.

Demonstrated robustness against variations in machine sound signals.

Abstract

As the labor force decreases, the demand for labor-saving automatic anomalous sound detection technology that conducts maintenance of industrial equipment has grown. Conventional approaches detect anomalies based on the reconstruction errors of an autoencoder. However, when the target machine sound is non-stationary, a reconstruction error tends to be large independent of an anomaly, and its variations increased because of the difficulty of predicting the edge frames. To solve the issue, we propose an approach to anomalous detection in which the model utilizes multiple frames of a spectrogram whose center frame is removed as an input, and it predicts an interpolation of the removed frame as an output. Rather than predicting the edge frames, the proposed approach makes the reconstruction error consistent with the anomaly. Experimental results showed that the proposed approach achieved 27% improvement based on the standard AUC score, especially against non-stationary machinery sounds.