• riconoscimento gesti
• virtuale
• gesture recognition
• hidden markov models
• fuzzy logic
• neural networks
• virtual reality

This thesis describes the design and the development of a recognition
system for human gestures. The main goal of this work is to demonstrate
the possibility to extract enough information, both semantic and quantitative,
from the human action, to perform complex tasks in a virtual environment.
To manage the complexity and the variability adaptive systems are
exploited, both in building a codebook (by unsupervised neural networks),
and to recognize the sequence of symbols describing a gesture (by Hidden
Markov models).