Hand Gesture Recognition and Virtual Game Control Based on 3D Accelerometer and EMG Sensors

My comments on other blogs:


Authors:
Xu Zhang - University of Science & Technology of China
Xiang Chen - University of Science & Technology of China
Wen-hui Wang - University of Science & Technology of China
Ji-hai Yang - University of Science & Technology of China
Vuokko Lantz - Nokia Research Center
Kong-qiao Wang - Nokia Research Center

Summary:
Accelerometers have been used as devices to accurately measure large movements but lack the precision needed to detect finger movement.  Electromyogram sensors have also been used to detect small subtle movements but cannot accurately measure large scale movements.

This paper introduces a hand based gesture device that combines both accelerometers (ACC) with electromyogram (EMG) sensors with the help of Hidden Markov Models (HMM) to increase accuracy.

A user study was done to test the effectiveness of ACC devices alone, EMG devices alone, and the combination of ACC+EMG devices.  From this study it showed that:

• ACC devices are only accurate 85.5-90.7% of the time
• EMG devices are only accurate 65.9-80.3% of the time
• ACC+EMG devices are accurate almost 100% of the time

The user study implemented a rubik's cube game that used this ACC+EMG device as input.

The participants were given a randomly mixed rubik's cube and asked to solve it with the device.  The user's were also given some common strategies to solve the rubik's cube quickly.

Discussion:
I thought this paper was definitely interesting although a little too much technical jargon was thrown that was pretty confusing.  Also, a lot of mathematical formulas were added and I was not able to fully understand what symbol's meant what.  From the pictures and data, it seems that this is a very plausible implementation of technology and is worth using in future research.  I think to fully trust the "almost 100% accuracy" reading I would have to test it myself.