ETHNICITY CLASSIFICATION FROM FACIAL IMAGES USING DEEP LEARNING METHODS

Abstract

Ethnicity classification from facial images is a vital task with broad applications in domains such as mental health diagnostics, demographic analysis, human-computer interaction, and social behavior modeling. This study presents a deep learning-based framework for robust ethnicity classification by leveraging facial expression features. The proposed method employs a Convolutional Neural Network (CNN) to extract deep spatial features from facial images, enabling the recognition of seven key facial expressions—happy, sad, neutral, anger, disgust, fear, and surprise—as intermediate attributes contributing to ethnicity identification. Extensive experiments were conducted on publicly available benchmark datasets, demonstrating superior performance over existing techniques. The model, trained using an unsupervised deep feature learning approach, achieved a high average classification accuracy of 95%, indicating its effectiveness and generalizability. The findings suggest that parsing facial expressions can significantly enhance the accuracy of ethnicity classification and support further research in affective computing and biometric identification.