Action recognition using machine learning techniques for robots

• Main Author: Yue, Zhongqi
• Other Authors: Huang Guangbin
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/72346
• Institution: Nanyang Technological University

Recent advancement in the processing capabilities of mobile chips has opened up the possibility of developing domestic companion robots with small form factors. The most intuitive way to interact with such robot is through human action, especially hand gesture. In this project, a robust static and dynamic gesture detector is built to achieve real-time performance on a mobile processor. The proposed framework features a novel hand hypotheses generator based on color and edge, a hand detector using Convolutional Neural Network (CNN), a static gesture recognizer based on skin contour analysis, and a hypotheses-tracking system based on Kalman Filter for increased performance and consistency. The resulting system is robust to viewpoint, ambient lighting, and rotation, capable of producing accurate results in various real-life settings. Index Terms: Hand Gesture Recognition, Hand Detection, Hand Tracking, Convolutional Neural Network