Using Deep Learning to Recognize Therapeutic Effects of Music Based on Emotions.

Area of Science:

• Computational neuroscience
• Music therapy
• Machine learning

Background:

• Music therapy is vital for health issues, with music listening as a key technique.
• Existing research lacks depth on music features and patient effects.
• Intelligent systems are needed to assist music therapists in song selection.

Purpose of the Study:

• To identify and predict the therapeutic benefits of music.
• To develop a machine learning model for personalized music recommendations in therapy.
• To bridge the gap between music features and their therapeutic impact on patients.

Main Methods:

• A multi-class neural network with three layers was developed for emotion classification and prediction.
• K-Fold Cross Validation was employed to evaluate the machine learning model's performance.
• The model was trained on a subset of the Million Dataset, incorporating musical, emotional, and solfeggio frequency features.

Main Results:

• The machine learning model demonstrated high performance in predicting therapeutic music effects.
• The system successfully classifies emotions and predicts music's suitability for individual users.
• User input on music preference and mood enables personalized song recommendations.

Conclusions:

• The developed machine learning system effectively predicts therapeutic music benefits.
• This intelligent system can assist music therapists and patients in selecting appropriate music for treatment.
• The model shows promise for enhancing music therapy outcomes through personalized recommendations.