2021/06/01 by Nik Dennler, Germain Haessig, Dennler, Nik +5 · 1 citation
Computer Science · Engineering · Neuroscience · #Advanced Memory and Neural Computing #Artificial Intelligence (cs.AI) #Audio and Speech Processing (eess.AS) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Neural Networks and Reservoir Computing #Neural and Evolutionary Computing (cs.NE) #Neural dynamics and brain function #Sound (cs.SD) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2106.00687
openalex publication_date 2021/06/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Vibration patterns yield valuable information about the health state of a\nrunning machine, which is commonly exploited in predictive maintenance tasks\nfor large industrial systems. However, the overhead, in terms of size,\ncomplexity and power budget, required by classical methods to exploit this\ninformation is often prohibitive for smaller-scale applications such as\nautonomous cars, drones or robotics. Here we propose a neuromorphic approach to\nperform vibration analysis using spiking neural networks that can be applied to\na wide range of scenarios. We present a spike-based end-to-end pipeline able to\ndetect system anomalies from vibration data, using building blocks that are\ncompatible with analog-digital neuromorphic circuits. This pipeline operates in\nan online unsupervised fashion, and relies on a cochlea model, on feedback\nadaptation and on a balanced spiking neural network. We show that the proposed\nmethod achieves state-of-the-art performance or better against two publicly\navailable data sets. Further, we demonstrate a working proof-of-concept\nimplemented on an asynchronous neuromorphic processor device. This work\nrepresents a significant step towards the design and implementation of\nautonomous low-power edge-computing devices for online vibration monitoring.\n