SELF-SUPERVISED ACOUSTIC ANOMALY DETECTION VIA CONTRASTIVE LEARNING

We propose an acoustic anomaly detection algorithm based on the framework of contrastive learning. Contrastive learning is a recently proposed self-supervised approach that has shown promising results in image classification and speech recognition. However, its application in anomaly detection is underexplored. Earlier studies have demonstrated that it can achieve state-of-the-art performance in image anomaly detection, but its capability in anomalous sound detection is yet to be investigated. For the first time, we propose a contrastive learning-based framework that is suitable for acoustic anomaly detection. Since most existing contrastive learning approaches are targeted toward images, the effect of other data transformations on the performance of the algorithm is unknown. Our framework learns a representation from unlabeled data by applying audio-specific data augmentations. We show that in the resulting latent space, normal and abnormal points are distinguishable. Experiments conducted on the MIMII dataset confirm that our approach can outperform competing methods in detecting anomalies.

SELF-SUPERVISED ACOUSTIC ANOMALY DETECTION VIA CONTRASTIVE LEARNING

We propose an acoustic anomaly detection algorithm based on the framework of contrastive learning. Contrastive learning is a recently proposed self-supervised approach that has shown promising results in image classification and speech recognition. However, its application in anomaly detection is underexplored. Earlier studies have demonstrated that it can achieve state-of-the-art performance in image anomaly detection, but its capability in anomalous sound detection is yet to be investigated. For the first time, we propose a contrastive learning-based framework that is suitable for acoustic anomaly detection. Since most existing contrastive learning approaches are targeted toward images, the effect of other data transformations on the performance of the algorithm is unknown. Our framework learns a representation from unlabeled data by applying audio-specific data augmentations. We show that in the resulting latent space, normal and abnormal points are distinguishable. Experiments conducted on the MIMII dataset confirm that our approach can outperform competing methods in detecting anomalies.

OutlierNets: Highly Compact Deep Autoencoder Network Architectures for On-Device Acoustic Anomaly Detection

Human operators often diagnose industrial machinery via anomalous sounds. Given the new advances in the field of machine learning, automated acoustic anomaly detection can lead to reliable maintenance of machinery. However, deep learning-driven anomaly detection methods often require an extensive amount of computational resources prohibiting their deployment in factories. Here we explore a machine-driven design exploration strategy to create OutlierNets, a family of highly compact deep convolutional autoencoder network architectures featuring as few as 686 parameters, model sizes as small as 2.7 KB, and as low as 2.8 million FLOPs, with a detection accuracy matching or exceeding published architectures with as many as 4 million parameters. The architectures are deployed on an Intel Core i5 as well as a ARM Cortex A72 to assess performance on hardware that is likely to be used in industry. Experimental results on the model’s latency show that the OutlierNet architectures can achieve as much as 30x lower latency than published networks.

Birth of a large volcanic edifice offshore Mayotte (Comoros Island, Western Indian Ocean)

<p>Volcanic eruptions are foundational events shaping the Earth’s surface and providing a window into deep Earth processes. We document here an ongoing magmatic event offshore Mayotte island (Western Indian Ocean) unprecedented in terms of emitted volume of lava and duration of the seismic crisis.This event gave birth to a deep-sea volcanic edifice 820m tall and ~ 5 km<sup>3</sup> in volume, located 50 km from Mayotte. A plume with distinct chemical signatures compared to open-ocean seawater emanated from the edifice, generating an exceptional 1900m-high vertical acoustic anomaly in the water column. Noble gas analyses in the vesicles from a popping rock dredged on the flank of the edifice, indicate rapid magma transfer from the asthenosphere. The edifice is located at the tip of a WNW-ESE–striking volcanic ridge composed of many other edifices, cones and lava flows constructed by past eruptions. Starting in May 2018 thousand of earthquakes were triggered by the magmatic event. The space-time distribution of the seismicity suggests that magma below the center of the ridge was transported to the new edifice over a few weeks in dikes that penetrated the brittle mantle a result of a lithosphere-scale extensional episode accommodating motion along a transfer zone between the East-African rifts and Madagascar. Since the eruption’s onset, the seismicity is mostly concentrated closer to the island, in an exceptionally deep zone (25-50 km) overlain by a zone of enigmatic, very low frequency, tremors.</p>

Time-frequency transforms in analysis of non-stationary quasi-periodic biomedical signal patterns for acoustic anomaly detection

Introduction: New approaches to efficient compression and digital processing of audio signals are relevant today. Thereis a lot of interest to new pattern recognition methods which can improve the quality of acoustic anomaly detection. Purpose:Comparative analysis of methods for time-frequency transformation of audio signal patterns, including non-stationary quasiperiodicbiomedical signals in the problem of acoustic anomaly detection. Results: The study compared different time-frequencytransforms (such as windowed Fourier, Gabor, Wigner, pseudo Wigner, smoothed pseudo Wigner, Choi — Williams, Bertrand, pseudoBertrand, smoothed pseudo Bertrand, and wavelet transforms) based on systematization of their functional characteristics(such as the existence and limitedness of basis functions, presence of zero moments and biorthogonal form, opportunity oftwo-dimensional representation and inverse transformation, real time processing, time-frequency transform quality, controlof time-frequency definition, time and frequency interference suppression, relative computational complexity, fast algorithmimplementation) for the problem of biomedial signal pattern recognition. A comparative table is presented with estimates ofinformation capacity for the considered time-frequency transforms. Practical relevance: The proposed approach can solve someacoustic anomaly detection algorithm implementation problems common in non-stationary quasi-periodic processes, in order tostudy disruptive effects causing a change in the functional state of ergatic system operators.