Decoding Mental Health: Speech Emotion Recognition in Detection Using Convolutional Neural Network

The integration of speech emotion recognition (SER) into psychiatric practice holds significant potential for advancing mental health diagnostics by enabling the classification of negative emotions closely linked to psychological disorders. Emotional states, particularly negative ones such as fear, anxiety, anger, and disgust, are strongly correlated with mental health conditions. Scientific evidence indicates that vocal characteristics undergo measurable changes with shifts in emotional states, offering a valuable tool for psychiatric evaluation. This study leverages the RAVDESS and TESS audio datasets to implement a Convolutional Neural Network (CNN) model, utilizing Mel-Frequency Cepstral Coefficients (MFCC) for feature extraction. Our model demonstrates robust performance in identifying a wide spectrum of emotions, achieving an impressive classification accuracy of 94%. This innovation underscores the potential of SER technology as a complementary tool in psychiatric assessment, augmenting traditional clinical methods and paving the way for more precise and objective diagnosis of mental health conditions.