Automatic Arabic Dialect Identification Using Deep Learning
Nejib Tibi, Mohamed Anouar Ben Messaoud · 2022
The language characteristics of a given community are referred to as dialect. Automatic speech recognition, e-health, and other applications benefit from being able to recognize dialect effectively. As a result, dialect identification has grown in importance and popularity as a study area. In this paper, we present an approach for automatic Arabic dialect identification based on language recognition system. It consists to apply the fundamental frequency, the energy, and the mel frequency cepstrum coefficients parameters based on multi-scale product analysis. These parameters are applied as input features to a convolutional neural network architecture for the classification of the dialect of speech signal. We can observe the efficiency of the application of multi-scale product to determine the characteristics of dialect. Our approach is evaluated on five dialect data set. The Experimental results show that the proposed approach outperforms several state-of-the-art methods, and is capable to identify the dialect.