GPGPU Implementation for Steering Hands Tracking of a Car Driver by Particle Filter with Depth Image Sensor

Abstract

An elaborated parallel implementation of particle filter in GPGPU has been applied to track steering hands of a car driver with depth image sensor for extracting arm and hand regions. Use of depth image sensor can provide more clear extraction of arm and hand regions for left and right than conventional image sensor. Parallel implementation in GPGPU allows fast computation and use of significantly larger number of particles in particle filter algorithm. Experimental results show computational performance of the proposed method.