SPEAKER ACCENT RECOGNITION USING MACHINE LEARNING

Speaker Accent Recognition using Machine Learning classifies speech into six accent groups: Spanish, French, German, Italian, British English, and American English. The executed notebook uses 329 records containing twelve Mel-frequency cepstral coefficient (MFCC) features and no missing values, together with ten WAV files for demonstration. A pipeline standardises the twelve features and trains a class-balanced Random Forest classifier with 500 trees on a stratified 80:20 split. The model is evaluated on 66 samples and the notebook also extracts mean MFCC vectors from uploaded WAV files for interactive prediction. The notebook reports 83.33% accuracy, 86.80% weighted precision, 83.33% weighted recall, and 81.66% weighted F1-score. American English dominates the dataset, and French, Italian, and British English have lower recall. The result is therefore presented as an academic prototype, not as evidence of universal accent identification. Keywords: speaker accent recognition, MFCC, audio classification, random forest, librosa, machine learning.