Genetic Algorithms in Speech Recognition Systems

In speech recognition, the training process plays an important role. When a good training model for a speech pattern is obtained, this not only enhances the speed of recognition tremendously .but also improves the quality of the overall performance in recognizing the speech utterance. In general, there are two classic approaches for this development, namely Dynamic Time Warping (DTW) and Hidden Markov Model (HMM). In this article, Genetic Algorithm (GA) is applied to solve involved nonlinear, discrete and constrained problems for DTW .Because of the intrinsic properties of GA, the associated non trival K-best paths of DTW can be identified without extra computational cost. The obtained results show the important contribution of the genetic algorithms in temporal alignment through the increasingly small factor of distortion.