CNN-Enhanced Speech Emotion Recognition with Recommendations

Speech Emotion Recognition (SER) plays a pivotal role in enhancing the efficiency and effectiveness of human-computer interaction (HCI). Understanding and interpreting emotional cues from speech signals enable a more natural and intuitive exchange between humans and machines. This project focuses on leveraging datasets from diverse sources, including RAVDESS, TESS, SA VEE, and CREMA, to develop a robust 1D Convolutional Neural Network (CNN) model. Employing Mel-frequency cepstral coefficients (MFCC) features, the model extracts pertinent information from speech signals. After rig-orous training and testing, the presented model demonstrated notable accuracy metrics, reaching 99% accuracy on the training dataset and 96.50 % accuracy on the test dataset. These results underscore the model's adeptness in identifying and categorizing emotional states conveyed through speech. The high accuracy suggests the potential for practical applications in real-world scenarios, particularly in the realm of human-computer interfaces. Moreover, the developed SER model has promising implications for speech-based recommendations. By accurately discerning emotional nuances in speech, the model can serve as a foundation for creating recommendation systems that cater to users' emotional states. The integration of such emotionally aware systems can significantly enhance user experience and engagement. This capability opens avenues for personalized recommendations, tailoring suggestions based not only on user preferences but also on their current emotional context.