A dome-shaped interface embedded with low-cost infrared sensors for car-game control by gesture recognition

Abstract

This paper proposes a steering wheel like interface using infrared sensors suitable for in-car control, car-game control or any interface with spin or turn hand gesture. Most of the interfaces introduced to-date use touch, position/depth sensing using cameras or proximity sensors positioned in a 2-D configuration. The electronic screen used for touch interface requires the user to maintain contact with specific positions on the screen. In contactless interfaces the sensors or camera are placed in a planar configuration, and complex gestures like turns or twist is intensive signal analysis. In the proposed preliminary model we introduce a contactless gesture recognition design shaped as a dome to allow natural hand movement for turns and tested to control a virtual object mimicking the movement of a car-wheel. The system recognizes hand movements like forward (translated as acceleration), backward (deceleration/slow), steady-hold (cruise), lateral for braking, turns-clockwise (right turn of the wheel) and anti-clockwise (left-turn of the wheel)– using 9 low-cost IR sensors embedded in a dome-shaped structure. The convex shape reduces interferences from adjacent sensors to a significant extent and allows for capturing distinct gestures. The inclusion of the acceleration and braking action to be controlled by the hand movement is to test and reduce leg and hand reflexes difference in the human visuo-motor feedback response system. The Hidden Markov Model was used for 5 basic gestures deduced from the IR signal analysis. The first version of the system was tested on a 3D virtual wheel-like object simulating a car tire. Real-time user gesture data tested against this model gave an overall average accuracy of 88.01% for the five gestures The user gestures were timed and were in the range of 140-300 ms depending on the gesture sequence. Some of the limitations of the first version of the design being addressed are noisy signals to reduce errors in gesture recognition. Secondly we need to test this on a comprehensive driving simulation to collect empirical data on the adaptation of the hand movement to control braking and acceleration.