Machine Learning Driven Audio Signal Analysis For Automated Hate Speech Detection In Short-Form Social Media Videos

6 Apr
Posted on Author

Authors: Assistant Professor, Mr.Y.Manas Kumar,, Iragavarapu Sri Vishnu Chittha Priya, Pepakayala Bhuvaneswari, Nookala Sri Giridhara Nageswara Adity, Koduri D P V Sai Sruthi, Yalla Naya Samson

Abstract: The rapid growth of social media platforms has significantly increased the spread of harmful and offensive content, including hate speech. Short-form video platforms allow users to share content quickly, making it challenging to monitor and control abusive speech. While most existing hate speech detection systems rely heavily on textual analysis, many harmful expressions occur through spoken language in video content. This study presents a machine learning-based approach for detecting hate speech directly from audio signals extracted from short-form online videos. The proposed framework collects audio data from publicly available social media videos and processes the signals using several audio feature extraction techniques such as Mel Frequency Cepstral Coefficients (MFCC), Spectral Centroid, Spectral Rolloff, Spectral Bandwidth, Zero Crossing Rate, and Chroma features. These features are used to train supervised machine learning models including Logistic Regression, Support Vector Machine, and Random Forest classifiers. To ensure reliable evaluation, a 5-fold cross-validation strategy is employed along with performance metrics such as accuracy, precision, recall, and F1-score. Experimental results demonstrate that the Random Forest model achieves superior performance compared to other classifiers by effectively capturing important audio characteristics associated with hate speech patterns. The study highlights the significance of spectral features and MFCC representations in identifying hateful expressions in speech. The proposed approach provides a practical framework for automated monitoring of harmful audio content in modern social media platforms and can contribute to improving online content moderation systems.

DOI: http://doi.org/10.61463/ijset.vol.14.issue2.180