Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 45189 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Sameti, Hossein
- Abstract:
- Keyword spotting with high speed and accuracy is an important subject whithin speech processing domain especially when we are dealing with various transmission channels. In this research discriminative keyword spotting methods are compared with HMM-based approaches. We have employed the discriminative approaches as our baseline methods due to their higher accuracy. The drawback of the conventional discriminative methods is their high computation cost and long execution time. The discriminative approach consists of two steps: feature extraction and classification. We have proposed four ideas to improve the performance of the baseline method. To improve the speed of the process, in feature extraction step phoneme classification based on GMM is proposed and for scoring the phoneme boundaries, using STM is suggested. For the classification step, a new method based on a full search for training the feature weights is presented. The discriminative keyword spotting needs a separate threshold for each keyword; in this research we have proposed two novel methods to specify the keyword thresholds, global and local methods. With our fast baseline keyword spotter, the accuracy is dropped by 4% with a 23X increase in computation speed. Using our proposed methods for feature extraction and weight training, the keyword spotting accuracy of the fast baseline method is increased by 2.78%. In addition, local thresholding results in 3% higher accuracy compared to the global thresholding algorithm
- Keywords:
- Keyword Spotting ; Thresholding ; Hidden Markov Model ; Phoneme Classifier ; Phoneme Boundary Function
Digital Object List-
محتواي کتاب - view
Bookmark