How do neural networks analyse drowsiness level for drivers?

Presented by Prof Helen Huang and Dr Ruihong Qiu

The social impacts of road crashes can be devastating for communities, and the economic costs are very high. Risks for serious crashes are increased in rural areas, for drivers of lower SES, and for drowsy drivers. At the same time, there has been an increase in unregulated transport work, potentially exacerbating differential risks for harm. Detection and prevention of risks is key to reducing this socio-economic burden.  In this project, we built a drowsiness detection model to predict driver drowsiness from simple devices. Different devices and sensors used in the laboratory are hard to deploy in a car in driving situations. Thus, a measurement available in both laboratory and real driving, heart rate data, was investigated. Heart rate data and drowsiness were measured simultaneously in the laboratory to serve as the training data. Once a drowsiness prediction model is built based on the heart rate data, it is expected to work in real-world context when the heart rate can be captured by common devices such as a smartwatch. In this project, we focus on developing neural network models that can accurately predict drowsiness given the heart rate. By applying advanced machine learning architectures such as multi-layer perceptron, recurrent neural network, gated recurrent unit etc., the developed predictive model can achieve encouraging performance. Yet there are more challenging scenarios to tackle in the future. This approach has applications for other forms of unregulated, precarious, and safety-critical work.

Prof Helen Huang is a Life Course Centre Chief Investigator at The University of Queensland. She is the Discipline Leader for Data Science at the School of Information Technology and Electrical Engineering. Her research spans across big data management, responsible data science, multimedia and social data analysis and machine learning, with more than 200+ papers published and ca 11,000 citations.

Dr Ruihong Qiu is a Life Course Centre Research Fellow at The University of Queensland. His research work on sequential modelling in user interactions has attracted interest from both academia and industry. His research has also generated broader impact to other disciplines, which has formed research collaborations with the Australian Centre of Water and Environmental Biotechnology and the School of Chemical Engineering.

Date & Time

Tue, 28 February, 2023

1:00 pm – 2:00 pm (AEDT)