Project information
- Category: Pattern Recognition
- Language: Python
- Project date: December, 2021
-
Video Source:
https://github.gatech.edu/omscs6601/assignment_6
https://omscs.gatech.edu/cs-6601-artificial-intelligence
Hidden Markov Models
The goal of this project was to build an American Sign Language (ASL) word recognizer, using measurments captured from ASL research videos. The measurments represented the Y-coordinates of the speaker's right hand. These measurments were used to build hidden Markov models (HMM's), representing the probabilistic pattern of a given word. The model was trained using data from videos of a speaker saying "buy", "house", and "car". Building the models involved calculating the transition probabilities and Gaussian distribution of each state, for each word. Finally, Viterbi Trellises were constructed and used to predict the likelihood of a given word by identifying the shortest traversal path between states.