(Research Fellow at the Dept. of Information Engineering, Università Politecnica delle Marche, Ancona, Italy)
DEEP LEARNING FOR ROAD CLASSIFICATION
Listening to road noise can be annoying. However, we can train machines to recognize the type of surface and the presence of wet in a glimpse from the tyre-road noise to make next-generation cars safer.
ROAD SURFACE DETECTION
Detecting the road roughness and wetness in intelligent vehicles allows to improve safety and sound quality inside the cabin.
This section reports resources related to this work.
WETNESS DETECTION
Being able to detect wet on the road allows to take safety measures in slippery conditions. We collected a large dataset of recordings from several microphone positions with wet and dry roads. We run CNN and BLSTM neural network on this corpus and obtain results as high as 95% and 97% respectively (F1-score).
Paper:
"Detecting Road Surface Wetness using Microphones and Convolutional Neural Networks", G. Pepe, L. Gabrielli, L. Ambrosini, S. Squartini, L. Cattani, in proc. 146th Audio Engineering Society Convention, Dublin 2019
SMOOTH/ROUGH CLASSIFICATION
The road roughness influence the grip and the rolling noise. Convolutional Neural Networks and a Siamese architecture allow high classification performance even without retraining the network on unseen tire types.
Papers:
"Deep Neural Networks for Road Surface Roughness Classification from Acoustic Signals", L. Ambrosini, L. Gabrielli, F. Vesperini, S. Squartini, L. Cattani, in proc. 144th Audio Engineering Society convention, Milan 2018
"Processing Acoustic Data with Siamese Neural Networks for Enhanced Road Roughness Classification", L. Gabrielli, L. Ambrosini, F. Vesperini, V. Bruschi, S. Squartini, L. Cattani, in proc. International Joint Conference on Neural Networks 2019 (IJCNN2019), Budapest
