US Accent Recognition Using Machine Learning Methods

Speech is considered the most important instrument for individuals to express their ideas and thoughts, and language is its most basic part. Even among speakers of the same language, there may be differences in language pronunciation. These differences are called accents. English is considered the most spoken language in the world. For this reason, there are many English speech accents. This study aims to recognize the American accent among six accents: American, British, French, German, Italian, and Spanish. Two different machine learning-based algorithms are employed for the accent recognition task: Support Vector Machine (SVM) and k-Nearest Neighbor (k-NN). In both methods, the user-defined hyperparameters are optimized to obtain high-accuracy results. Furthermore, the k-fold cross validation technique is applied to ensure the reliability of the results. Experimental results show that SVM with the Radial Basis Function (RBF) kernel yields the highest accuracy (89 %). As a result, SVM with RBF kernel is a viable option for speaker accent recognition applications.