A Machine Learning Approach to Recognize Speakers Region of the United Kingdom from Continuous Speech Based on Accent Classification

Abstract

Speech is one of the primary modes of communication with a lot of identical features for measuring performance and behavior of human voice. Accent is an important element and can play a vital role in spoken language. In this paper, we propose a region detection approach of UK citizens by recognizing their accent from continuous speech. The ultimate goal of this paper is to detect the region of UK citizens from which region among Ireland, Midland, Northern England, Scotland, Southern England and Wales he/she belongs using continues speech. Firstly, we use Mel Frequency Cepstral Coefficient (MFCC) for extracting the feature from continuous speech. Then we applied several Machine Learning classifiers to train and test our model. After evaluating performance we find that k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), and Random Forest classifier provide comparatively better accuracy than others. We also perform a comparative analysis of these three algorithms. We got the best accuracy of 98.48% by applying k-NN classifier.