Abstract
<title>Abstract</title> <p>Accent detection plays a crucial role in consumer electronics by enabling voice authentication in devices such as smart TVs, remote controls, smart home assistants, and hands-free systems. Despite its significance, accent recognition remains a challenging task due to linguistic variability and acoustic diversity. This article presents a comprehensive comparative study of machine learning (ML) and deep learning (DL) approaches for speech accent recognition across multilingual datasets. The Speech Accent Archive (SAA) and AccentDB datasets are utilized to evaluate English and multiple Indian language accents. A diverse set of acoustic features, including Mel-frequency cepstral coefficients (MFCCs), first- and second-order derivatives, Mel-spectrograms, and chromagram representations, is systematically extracted and analyzed. Multiple ML classifiers—k-nearest neighbors (k-NN), support vector machines (SVM), stochastic gradient descent (SGD), random forests (RFs), AdaBoost, gradient boosting (GB), XGBoost, and multilayer perceptron (MLP)—are implemented and evaluated alongside DL architectures such as convolutional neural networks (CNN), bidirectional long short-term memory (BiLSTM), and hybrid CNN–BiLSTM models. Performance is assessed using precision, recall, F1-score, accuracy, and loss metrics. Experimental results demonstrate that for the SAA dataset, MLP, GB, and AdaBoost achieve accuracies above 84%, while for the AccentDB dataset, MLP (98.37%), SVM (98.08%), and SGD (96.82%) exhibit exceptional performance. Among DL approaches, CNN-based architectures consistently yield competitive results across both datasets, achieving the highest accuracies of 91.47% on the SAA dataset and 99.18% on the AccentDB dataset. The key contribution of this work lies in its unified evaluation of a broad spectrum of ML and DL models across heterogeneous accent datasets, offering practical insights into model selection for real-world deployment. The findings highlight the feasibility of integrating efficient accent recognition models into consumer electronics to enable personalized, and accent-aware user interactions.</p>