In the domains of music knowledge extraction and sound psychological research, emotional evaluation of songs has been a hot topic. Music has been shown to provoke a wide range of emotions in people, which can have a positive impact on their mental and physical well-being. Music has long been known for its strong emotional impact and potential therapeutic benefits. Deep learning algorithms have lately shown great promise in a wide range of applications, including music emotion analysis. The study collected a large dataset, which included music sound recordings and personality assessments. The study extracts aspects from music compositions such as pitch, rhythm, timbre, and loudness. These features are then used to create a convolutional neural network (CNN) deep learning model. The CNN is trained using a collection of music recordings labeled with one of six emotions: happy, sad, angry, startled, fearful, or neural. The findings show that the proposed approach achieved an accuracy of 80% in recognizing the feelings of songs for therapeutic purposes and may be used to find music compositions that are considered to be beneficial for people with a variety of emotional needs. The research indicates that machine learning could operate as a useful instrument for melodious therapies, and thus proposes integrating cutting-edge artificial intelligence (AI) with the psychological effect of melodies to reveal new possibilities for personalized therapy and support a deeper understanding of the intricate link between sounds and feelings.