Gesture-Based Control in a Smart Home Environment

Abstract

Gesture is a convenient and natural way to control a smart home. The wearable device provides an excellent vehicle for getting a user's hand gesture. Recognition models for gestures can be divided into two types: user dependent and user independent. In this research, we propose a hybrid model that combines both user dependent and user independent models to distinguish a user's hand gestures. Our research investigates which model among three is the best approach for recognizing hand gestures. We employ ten hand gestures as the test cases for comparison. First, from a 6-axis wearable device we extract features based on the collected raw data of hand gestures. Then these extracted features are analyzed by a oneM2M-compliant platform to detect gestures based on Decision Tree and Logistic Regression algorithms. With a data set of over 7 users and 20 repetitions of tests for each user, we tested the effectiveness of recognition models and gesture detection algorithms. The results show that our proposed hybrid model could achieve the best accuracy with either of two detection algorithms.