HMM parameters estimation using hybrid Baum-Welch genetic algorithm
Mourad Oudelha, Raja Noor Ainon · 2010
Automatic speech recognition (ASR) has been for a long time an active research area with a large variety of applications in human-computer intelligent interaction. The most used technique in this field is the Hidden Markov Models (HMMs) which is a statistical and extremely powerful method. The HMM model parameters are crucial information in HMM ASR process, and directly influence the recognition precision since they can make an excellent description of the speech signal. Therefore optimizing HMM parameters is still an important and challenging work in automatic speech recognition research area. Usually the Baum-Welch (B-W) Algorithm is used to calculate the HMM model parameters. However, the B-W algorithm uses an initial random guess of the parameters, therefore after convergence the output tends to be close to this initial value of the algorithm, which is not necessarily the global optimum of the model parameters. In this paper, a Genetic Algorithm (GA) combined with Baum-Welch (GA-BW) is proposed; the idea is to use GA exploration ability to obtain the optimal parameters within the solution space.