Clemson University
Investigation of Training Algorithms for Hidden Markov Models Applied to Automatic Speech Recognition
Abstract
dc:description.abstractThe work presented in this thesis focuses on simulating a speech recognizer which is trained by different people with different speaking styles and investigates how sensitive the training and recognition processes are to the variations in the training data. There are four main parts to this work. The first involves an experiment of weighting methods for training with multiple observation sequences. The second involves the testing of different initial parameters. The third part includes the first experiment involving training with multiple observation sequences. The model's sensitivity to variations in training data was evaluated by comparing the cases of different values of variation. The final part varied the observation vectors with the variation restricted to only one of the eight positions in the sequence. The experiment was repeated for each of eight positions in the observation sequence, and the effect on recognition was evaluated.
Degree
thesis:*- Name thesis:degree_name
- Master of Science (MS)
- Level thesis:degree_level
- Thesis
- Discipline thesis:degree_discipline
- Computer Engineering
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Fang, Eric
- Contributors dc:contributor
-
- Gowdy, John
- Schalkoff , Robert
- Birchfield , Stanley
Subjects
dc:subject × 4Identifiers
dc:identifier.*- Repository record dc:identifier
- https://open.clemson.edu/all_theses/580