A hybrid neural network for Arabic Internet navigator commands recognition

A hybrid neural network architecture called OMART2-FAM is introduced. It consists of two neural networks connected by an intermediate memory. Optimized match adaptive resonance theory (OMART2) neural network and fuzzy associative memory (FAM) neural network are used for Arabic phoneme signals and Arabic word signals recognition respectively. The intermediate memory is a feedforward field, which retains and encodes the sequence of recognized phonemes at F2 output field of OMART2. Implementing complement coding normalizes connections between the intermediate memory and the input field of FAM. OMART2-FAM classifier of Arabic Internet navigator command signals is implemented. Experimental results show that the new algorithm of OMART2 generally exhibits faster learning and better clustering performance. Additionally, they illustrate the effectiveness of the new hybrid neural network OMART2-FAM; hence the justification for its implementation in a speech recognition system of Arabic commands is to maximize generalization and minimize misclassification error rates.