Hand Gesture Recognition System Utilizing Hidden Markov Model for Computer Visions Applications

Isa Ibrahim
Email: isibr1@morgan.edu
Electrical and Computer engineering Dept.
Morgan State University, Baltimore
Maryland, USA

Abstract

This work describes a hand gesture recognition system utilizing Hidden Markov Models (HMMs) for computer vision applications. The system processes video input of hand gestures through skin color-based segmentation, morphological operations, hand segmentation, and hand tracking and trajectory smoothing. The HMM with Gaussian emissions is implemented using the HMM learn package, and the Viterbi algorithm is used to decode an observation sequence and determine the most likely state sequence and its probability. The work also presents the methodology for data collection, preparation, and augmentation, as well as the quantization of discrete observations and Baum-Welch re-estimation algorithm. The performance of the system is evaluated using a test set of observation sequences, and the accuracy of the maximum likelihood classifier for recognizing letters is assessed using a validation set. The results demonstrate the effectiveness of the system for accurately recognizing hand gestures and corresponding letters.

Keywords: HMM- Hidden Markov Model; CNN – Convolutional Neural Networks; RGB – Red, Green, Blue; ROI – Region of Interest