Correct Pronunciation Detection of the Arabic Alphabet Using Deep Learning

Automatic speech recognition for Arabic has its unique challenges and there been relatively slow progress in this domain. Specifically, Classic received even less research attention. The correct pronunciation of the alphabet significant implications on meaning words. In work, we have designed learning models classification based an alphabet. is a challenging task community. We divide problem into two steps, firstly train model to recognize alphabet, namely classification. Secondly, determine quality pronunciation, Due availability audio data kind, had collect from experts, novices our model’s training. To these models, extract features using mel-spectrogram. employed deep convolution neural network (DCNN), AlexNet with transfer learning, bidirectional long short-term memory (BLSTM), type recurrent (RNN), data. For classification, DCNN, AlexNet, BLSTM achieve accuracy 95.95%, 98.41%, 88.32%, respectively. 97.88%, 99.14%, 77.71%,