Abstract:In this paper, we propose an anomaly detection algorithm for machine sounds with a deep complex network trained by self-supervision. Using the fact that phase continuity information is crucial for detecting abnormalities in time-series signals, our proposed algorithm utilizes the complex spectrum as an input and performs complex number arithmetic throughout the entire process. Since the usefulness of phase information can vary depending on the type of machine sound, we also apply an attention mechanism to control the weights of the complex and magnitude spectrum bottleneck features depending on the machine type. We train our network to perform a self-supervised task that classifies the machine identifier (id) of normal input sounds among multiple classes. At test time, an input signal is detected as anomalous if the trained model is unable to correctly classify the id. In other words, we determine the presence of an anomality when the output cross-entropy score of the multiclass identification task is lower than a pre-defined threshold. Experiments with the MIMII dataset show that the proposed algorithm has a much higher area under the curve (AUC) score than conventional magnitude spectrum-based algorithms.

What problem does this paper attempt to address?

The paper is primarily dedicated to addressing the issue of machine sound anomaly detection. Specifically, the authors propose a complex network algorithm based on self-supervised learning to detect anomalies in machine sounds. This method leverages a deep complex network and particularly focuses on the importance of phase continuity information in time series signal anomaly detection. The key contributions of the paper include: 1. **Introduction of complex spectrum input**: Traditional anomaly detection methods typically use only the magnitude spectrum as feature input. In contrast, the algorithm proposed in this paper uses the complex spectrum as input and performs complex operations throughout the process to fully utilize phase information. 2. **Application of attention mechanism**: Considering that the effectiveness of phase information may vary for different types of machine sounds, the algorithm introduces an attention mechanism to control the importance weights of complex spectrum and magnitude spectrum features based on different machine types. 3. **Self-supervised learning task**: The network is trained to classify the machine identity of normal input sounds, i.e., a self-supervised classification task. During the testing phase, if the trained model fails to correctly classify the machine identity of the input signal, it is considered an anomaly. 4. **Experimental validation**: Experiments were conducted using the MIMII dataset, and the results show that the proposed algorithm has a significant advantage in terms of the Area Under the Receiver Operating Characteristic Curve (AUC) score compared to traditional methods that are based solely on the magnitude spectrum. In summary, this research aims to develop a new method capable of effectively detecting anomalies in various types of machine sounds, particularly by combining self-supervised learning techniques and complex network architectures to achieve this goal.

Self-supervised Complex Network for Machine Sound Anomaly Detection

Vibration Anomaly Detection of Wind Turbine Based on Temporal Convolutional Network and Support Vector Data Description

Machine Anomalous Sound Detection Based on Self-Supervised Classification

Semi-Supervised Machine Condition Monitoring by Learning Deep Discriminative Audio Features

Low-complexity Attention-based Unsupervised Anomalous Sound Detection exploiting Separable Convolutions and Angular Loss

Exploring Large Scale Pre-Trained Models for Robust Machine Anomalous Sound Detection

Anomalous Sound Detection using Audio Representation with Machine ID based Contrastive Learning Pretraining

Representational learning for an anomalous sound detection system with source separation model

Anomalous Sound Detection Using Self-Attention-Based Frequency Pattern Analysis of Machine Sounds

Unsupervised Anomalous Sound Detection for Machine Condition Monitoring Using Classification-Based Methods

Domain Shift-oriented Machine Anomalous Sound Detection Model Based on Self-Supervised Learning

Spectro Temporal Fusion with CLSTM-Autoencoder based approach for Anomalous Sound Detection

Robust Audio Anomaly Detection

Outlier-aware Inlier Modeling and Multi-scale Scoring for Anomalous Sound Detection via Multitask Learning

Machine Anomalous Sound Detection Using Spectral-temporal Modulation Representations Derived from Machine-specific Filterbanks

A Dual-Path Framework with Frequency-and-Time Excited Network for Anomalous Sound Detection

Anomalous Sound Detection as a Simple Binary Classification Problem with Careful Selection of Proxy Outlier Examples

Unsupervised Detection of Anomalous Sound based on Deep Learning and the Neyman-Pearson Lemma

Deep-Learning-Based Approach to Anomaly Detection Techniques for Large Acoustic Data in Machine Operation

Unsupervised Anomaly Detection and Localization of Machine Audio: A GAN-based Approach

Regularized Contrastive Masked Autoencoder Model for Machinery Anomaly Detection Using Diffusion-Based Data Augmentation