Malay dialect identification using Bi-LSTM trained on MFCC features

The Malay language is a major language in the Austronesian family and is commonly spoken in various parts in Southeast Asia (SEA). Despite its many native speakers, research on intelligent techniques to analyse the language has been limited. In this paper, we present a Long Short-Term Memory (LSTM) to perform dialect recognition for the Malay Language. 240 samples were collected from 10 native dialect speakers to perform the experiments. Subsequently, we represented the raw audio recordings as Mel Frequency Cepstrum Coefficient (MFCC) features to train the LSTM classifier. The results achieved 98.20% classification accuracy, comparable to similar current methods.