We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Two layered gated recurrent stacked long short-term memory networks for driver's behavior analysis.
- Authors
Sahoo, Goutam Kumar; Das, Santos Kumar; Singh, Poonam
- Abstract
In this work, an in-vehicle deep learning-based driver assistance system is proposed to monitor driver behavior to reduce the risk of road traffic accidents. A two-layer gated recurrent stacked long short-term memory (LSTM) network is implemented to detect and classify driver behaviors such as normal, aggressive and drowsy driving. The methodology to develop the proposed model is based on a many-to-one architecture, where two LSTM layers (LSTM and Bi-LSTM) and two gated recurrent units (GRUs) are stacked with 32 hidden cells in each layer and one fully connected output layer. Python and the Keras-TensorFlow libraries are used to build various deep-learning models. A publicly available benchmark smartphone dataset, the UAH-DriveSet dataset, is used to evaluate the performance of the proposed model, which achieves good estimation with a driver behavior classification accuracy of 92%. The performance comparison showed superiority over state-of-the-art techniques when the cascaded deep-learning architecture was evaluated and validated in a simulation environment. This model can be used with an in-vehicle dashboard-mounted smartphone that is set to monitor driving behavior and alert the driver to maintain normal driving.
- Subjects
DEEP learning; BEHAVIORAL assessment; DRIVER assistance systems; MOTOR vehicle driving; AGGRESSIVE driving; CRANES (Birds); INTELLIGENT transportation systems
- Publication
Sādhanā: Academy Proceedings in Engineering Sciences, 2023, Vol 48, Issue 2, p1
- ISSN
0256-2499
- Publication type
Article
- DOI
10.1007/s12046-023-02126-y