Clemson University
A Speaker Independent Isolated Word Recognition System Using Linear Prediction
Abstract
dc:description.abstractThis dissertation describes a speaker independent isolated word recognition system using autoregression (linear prediction) on speech samples. A number of distance measures for speaker independent recognition of isolated words are proposed and evaluated for their accuracy of recognition. These distance measures use autocorrelation coefficients alone o r autocorre lation coefficients and linear predictor coefficients as feature parameters of the speech samples. Actual evaluation of these distance measures is then perfor1ned using a standard 40 word reading test vocabulary spoken by 2S differentspeakers. The best distance measure is further explored in conjunction with both the nearest neighbor and the Knearest-neighbor decision rules. A combined decision algorithm using both K-nearest-neighbor and nearest-neighbor algorithms is then formulated. Recognition results using the three different algorithms are then compared. It is observed that this combined decision algorithm gives better results than either the nearest-neighbor or the K-nearestneighbor decision rule used alone. The number of speakers was then increased to lOS to show the statistical significance of the results obtained in this project. The recognition rate obtained with the best procedure for lOS speakers was 89.2 percent. The recognition time for this procedure was 9.8 seconds per utterance.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy (PhD)
- Level thesis:degree_level
- Dissertation
- Discipline thesis:degree_discipline
- Electrical and Computer Engineering
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Gupta, Vishwa Nath
- Contributors dc:contributor
-
- John M. Sondy
- Maruice L. Holls
- K. L. Wallenins
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://open.clemson.edu/arv_dissertations/16