Hand Detection for Advanced Driver Assistance System
Tian Zhou, Preeti Pillai, Ganesh Yalla and Kentaro Oguchi Toyota InfoTechnology Center, Mountain View, CA Jan 2016 - May 2016
Description: Driver modeling is of great importance in Advanced Driving Assistance Systems (ADAS), it serves as one of the building blocks for many ADAS systems. For example, it can be used to model the normal driving behavior for each driver, so that abnormal activities (potential distraction) can be spotted. Additionally, the driver’s behavior can be leveraged to predict his driving activities (like braking or making a maneuver). Having an accurate and robust driver-modeling module would greatly help most ADAS systems.
This presentation will feature some of the progress made towards driver modeling conducted during Tian Zhou’s internship at ICD, Toyota ITC from Jan 2016 to May 2016. Firstly, a hand detection module specially tailored for driving context will be introduced. Both traditional approaches and deep learning approaches will be covered and compared. Secondly, a driver modeling algorithm will be presented. This algorithm leverages the hand activities and builds a probabilistic model to capture the driving habit of each driver. Lastly, some sample applications which are based on the driver modeling module will be described.