ABSTRACT
Machine learning techniques are widely used for accent classification. Due to the accent, the pronunciation differs, and that leads others to think of it as a different language. In this case, classifying the accents in a language helps identify it as a specific language. This paper identifies the Indian, American, and British English accents. Initially, the model processes the input speech signals, removes noise, and converts them into a format suitable for the Mel-Frequency Cepstral Coefficients (MFCCs) processing. And then, the features are extracted using the MFCCs. These extracted features are used to train the Hidden Markov Model (HMM) which uses labeled speech samples. The trained HMM model is tested and is used to predict the accent of an input speech sample. Most researchers are using the Convolution Neural Network (CNN) for classification. In order to improve the efficiency of the model, we are using HMM.
