Turn-Intent Analysis Using Body Pose for Intelligent Driver Assistance

Abstract

Pervasive computing technologies hold much promise for enhancing automotive safety by introducing a new range of human-centered driver assistance systems. Requirements for designing an active safety system are accurately, reliably, and quickly identifying the conditions leading to an accident and inducing corrective actions to prevent the accident. The authors propose a driver turn-maneuver prediction system using a two-class pattern classification algorithm using driver-pose and steering-angle information. They analyze classifier-detection performance using receiver-operator-characteristic curves. These curves provide a picture of the attainable proactivity versus transparency ratios, pertaining to a pervasive computing system's ability to foresee the user's needs as compared to the system's ability not to annoy the user. The goal is to motivate the development of both vision-based body-pose recovery and behavior recognition algorithms for driver assistance systems. This article is part of a special issue on Intelligent Transportation.