Radar based deep learning technology for loudspeaker faults detection and classification

Izzo, A. and Clemente, C. and Ausiello, L. and Soraghan, J.J.; (2020) Radar based deep learning technology for loudspeaker faults detection and classification. In: 2020 IEEE Radar Conference (RadarConf20). IEEE, ITA. ISBN 9781728189420 (https://doi.org/10.1109/RadarConf2043947.2020.9266...)

[thumbnail of Izzo-etal-IEEE-RC-2020-Radar-based-deep-learning-technology-for-loudspeaker-faults]
Preview
Text. Filename: Izzo_etal_IEEE_RC_2020_Radar_based_deep_learning_technology_for_loudspeaker_faults.pdf
Accepted Author Manuscript

Download (623kB)| Preview

Abstract

Recently, radar based micro-Doppler signature analysis has been successfully applied in various sectors including both defence and civilian applications. A joint radar micro-Doppler and deep learning technology for End-Of-Line (EOL)test of loudspeakers is proposed in this paper. This approach offers the potential benefits of characterizing the mechanical motion of a loudspeaker in a noisy environment as a production line, in order to automatically identify and classify defects. Starting from real radar signal, the proposed Bidirectional Long Short-Term Memory (BiLSTM) classifier has been tested on training, validation and test dataset. The results show that the proposed approach produces a probability of correct classification abovethe98%, outperforming the traditional k-NN classifier.