Date of Graduation

5-2015

Document Type

Thesis

Degree Name

Master of Science in Electrical Engineering (MSEE)

Degree Level

Graduate

Department

Electrical Engineering

Advisor

Vijay K. Varadan

Committee Member

Dr. Juan C. Balda

Second Committee Member

Dr. Roy McCann

Keywords

Applied sciences; Electrical engineering; Electromyography; Hand pattern; Pattern recognition

Abstract

The Importance of gesture recognition has widely spread around the world. Many research strategies have been proposed to study and recognize gestures, especially facial and hand gestures. Distinguishing and recognizing hand gestures is vital in hotspot fields such as bionic parts, powered exoskeleton, diagnosing muscle disorders, etc. Recognizing such gesture patterns can also create a stress-free and fancy user interface for mobile phones, gaming consoles and other such devices.

The objective is to design a simple yet efficient wearable hand gesture recognizing system. This thesis also shows that by taking both EMG and accelerometer data into account, can improve the system to recognize more patterns with higher accuracy levels. For this, a hand band embedded with a triple axis accelerometer and three surface EMG electrodes is employed to source the system. The non-invasive surface EMG electrodes senses muscle action while the accelerometer senses the hand motions. The EMG signal is passed through analog front-end module for noise filtering and signal amplification. An ARM Cortex processor converts the analog EMG and accelerometer signal into digital and transmits to a PC via Bluetooth protocol. On the receiver section, the raw EMG and acceleration data is further processed and decomposed offline using MATLAB tools to extract features such as root mean square, waveform length, threshold crossing, variance and mean. Extracted features are then fed through multi-class SVM (Support Vector Machine) process for pattern recognition. The chapters below discuss in greater detail on pattern recognition technique and other modules involved.

Share

COinS