The speaker identification based on their accent is a challenging task, especially for Indian regional dialects, due to the subtle metric variations and close phonetic similarities. The aim of this research work is to create automated speaker identification (ASI) system that can categorise native speakers based on their regional accents in North Indian languages: Hindi (HIN), Punjabi (PUN), and Bengali (BEN). The proposed method is to collect the speech data from individuals who speak these languages as their native language, thereafter Mel-frequency cepstral coefficients (MFCC) and linear predictive cepstral coefficients (LPCC) are used to extract the features. After feature extraction, machine learning (ML) classification techniques are used, such as support vector machines (SVM) and decision trees (DT), on non-native English speech samples influenced by native accents. Experimental results demonstrate that the SVM classifier significantly outperforms the DT classifier, achieving an accuracy of 96.23% compared to 93.75%. These findings suggest that cepstral features, combined with SVM classifiers, offer a robust approach for identifying native language speakers in regional Indian accents. The proposed system has potential applications in improving speech recognition, language learning, and biometric authentication within multilingual environments.
Singh et al. (Fri,) studied this question.