الفهرس Only 14 pages are availabe for public view
 |  | from | 90 |  |  |
| | | from | 90 | | |
Abstract
The thesis aims to compare between three different adaptation algorithms that , are: batch expectation maximization (EM) algorithm. incremental expectation maximization (EM) algorithm. and batch k-means algorithm. wing Arabic language.
Parameter estimation for a hidden Markov model (HMM) is performed using an expectation maximization (EM) algorithm with the maximum-likelihood (ML) criterion. The EM, algorithm is an iterative scheme that is well-defined and numerically stable. but convergence may require a large number of iterations. The multiple iterations required for the EM algorithm to converge make these adaptation schemes computationally expensive.
The incremental EM algorithm speed-UP convergence without any actual loss of recognition performance. The algorithm selects a subset of data from the observation set, update the model parameters based on the subset, and then iterates the process until convergence of the parameters or certain Dumber of iterations. Experimental results show that the incremental EM algorithm is substantially faster than the batch EM algorithm. and reduces the computational requirements SO it takes lower time than the batch EM.
The batch k-meansalgorithm is sometimes preferable to use instead of EM algorithms, because Viterbi decoding is widespread search technique: in HMM based speech recognizers. The experimental results show that the batch k-means algorithm has reached convergence faster than batch EM. There is no actual degradation in performance if batch EM is used than batch k-means. But batch k-means has more computational requirements, and takes more time than batch EM.