Comparison of Emotion Recognition from Baby Cries Using CNN and SVM with Noise Reduction Techniques

Emotion recognition from baby cries has become an important field of research, with potential applications in healthcare and caregiving. This paper presents a comparison of Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) for classifying emotions from baby cries, focusing on categories such as hunger, pain, tiredness, discomfort, and burping. Noise interference is a significant challenge in cry recognition, and noise reduction techniques are applied to enhance model accuracy. A 5-fold cross-validation approach is used to evaluate the performance of both models under noisy conditions. According to preliminary findings, the SVM outperformed the CNN, which had an accuracy of 81.7%, with an accuracy of 84% across all emotion categories. Both models demonstrated significant improvements with noise reduction, with the SVM model achieving the highest accuracy increase, rising from 81.2% to 84%. This paper highlights the effectiveness of deep learning models for real-time emotion detection and underscores the importance of noise reduction in achieving reliable results.