A Multi-Speaker System for Arabic Speech Perception
Mohamed Hamed, Dalia Wafik · 2018
This paper presents a new generalized system for the Arabic speech recognition based on 3-layer integrated neural networks consisting of control network and several sub-networks. The segmentation technique is accounted to minimize the selected features, and consequently, the computational time effort. The back propagation and the linear predictive coding bases are considered in the tested multi-speaker system. The proposed concept is applied for words representing the digits. The recognition rate is quite accurate, and so it is recommended to be used in the software suitable for Arabic speech recognition. The multi-speaker signals are tested and trained for recognition. They were classified into three groups as male, female and children groups. The effect of number of nodes in the hidden layer of the neural network on the rate of recognition is investigated.