Automatic Dialect Classification using SVM

Automatic Dialect Classification has attracted researchers in the field of speech signal processing. Dialect is defined as the language characteristics of a specific community. As such, dialect can be recognized by speaker phonemes, pronunciation, and traits such as tonality, loudness, and nasality. Dialect classification is a substantial tool in speech recognition and has the potential to improve the efficiency of Automatic Speech Recognition systems. This paper presents a study of different dialects in English language (American) and features that are useful for their classification. The experiment demonstrates that there are several features of the speech signal which are conducive for recognizing different dialects within a language such as chroma features and spectral features. Other speech features including MFCC and FDLP were also used with these features in order to improve the performance of the classifier. The supervised machine learning classifier that has been used in our research is the Support Vector Machine. Some refinements were introduced to the existing chroma feature extraction processes to make them more suitable for speech signal classification.