Coalition of genetic algorithms and artificial neural network for isolated spoken Malay speech recognition / Noraini Seman

• Main Author: Seman, Noraini
• Format: Book Section
• Language: English
• Published: Institute of Graduate Studies, UiTM 2012
• Subjects: Malaysia
• Online Access: http://ir.uitm.edu.my/id/eprint/19099/1/ABS_NORAINI%20SEMAN%20TDRA%20VOL%201%20IGS%2012.pdf; http://ir.uitm.edu.my/id/eprint/19099/

The automatic speech recognition (ASR) field has become one of the leading speech technology areas using artificial intelligence (AI) approaches. Despite all of the advances in the speech recognition area, the problem is far from being completely solved. Various methods have been introduced to develop an efficient ASR system. A variety of automatic knowledge acquisition or learning and adaptation concepts need to be established in speech recognition using AI approaches. These key concepts can only be implemented using artificial neural networks (ANNs) approach. However, traditional ANNs have many fundamental problems regarding a long and uncertain training process, which in most cases learning or training of a neural network is based on a trial and error method. Genetic Algorithm (GA) based learning technique provides an alternative way that involves controlling the learning complexity by adjusting the number of weights of the ANN. However, due to the stochastic nature of this algorithm, the learning process can reach an optimal solution with much higher probability than many standard neural network techniques.