AbstractThe increasing reliance on voice-activated technology in daily life has highlighted the need forASR systems that can accurately understand a range of English accents. Current systems of-ten struggle with diverse accents, impacting the inclusivity and effectiveness of speech-basedinteractions. This thesis explores accent classification in the English language, focusing onfive major accents: American, British, Indian, Australian, and Canadian. Using supervisedmachine learning methods, including CNN, SVM, and RF models, this study investigates theeffectiveness of these models in classifying different accents.The research uses the Mozilla Common Voice dataset, which offers a diverse collection oflabeled audio samples. Through experimental evaluation, the performance of CNN, SVM,and RF models is assessed in accurately classifying the specified accents. The findings in-dicate that while CNNs achieve an accuracy of 70%, traditional machine learning methodsoutperform CNNs, with SVMs reaching an accuracy of 83% and RFs achieving 79%. Notably,SVMs excelled in differentiating US and Canadian accents with precision rates of 90% and88%, respectively.Overall, this work contributes to the development of more inclusive and adaptable speechrecognition systems, benefiting applications in customer service, virtual assistance, and othervoice-based technologies. Additionally, the thesis reflects on the challenges encountered dur-ing the study, including model over-fitting and difficulties in distinguishing subtle accentvariations, and discusses potential avenues for future research to enhance accent classifica-tion methodologies.