Social Media Sentiments Analysis using Convolutional Neural Network and Support Vector Machine

Speech emotion recognition (SER) is a critical area of research in human-computer interaction, enabling applications such as sentiment analysis, virtual assistants, and psychological assessment. This study proposes a hybrid approach utilizing a 1D Convolutional Neural Network (CNN1D) and a Support Vector Machine (SVM) for classifying emotions from speech signals into seven categories: Angry, Disgust, Fear, Happy, Sad, Surprise, and Neutral. The Toronto Emotional Speech Set (TESS) and the Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) datasets were employed to ensure diverse emotional representations. The preprocessing phase involved silence removal to eliminate non-informative segments, followed by Independent Component Analysis (ICA) to enhance signal quality. Feature extraction techniques, including Mel Frequency Cepstral Coefficients (MFCCs), Zero Crossing Rate (ZCR), Chroma Short-Time Fourier Transform (Chroma STFT), Root Mean Square (RMS) energy, and Mel Spectrogram, were used to capture spectral, temporal, and tonal characteristics of speech. To improve model generalization, data augmentation techniques such as noise addition, time stretching, pitch shifting, and time shifting were applied. Experimental results demonstrate that the combined CNN-1D and SVM approach effectively captures emotional variations in speech, providing improved classification accuracy. This research contributes to the advancement of robust SER models by integrating deep learning with traditional machine learning, optimizing both feature representation and classification performance.