This study introduces a Music Emotion Recognition and Recommendation System driven by Electroencephalography (EEG) signals to detect emotional states and deliver personalized music experiences. EEG, which is recorded through scalp electrodes, is conditioned so as to reject artifacts originating from eye movement and muscle activity. In analyzing these cleaned signals, the more important frequency bands associated with present emotional and cognitive states, such as happiness, calmness, excitement, and melancholy are alpha, beta, theta, and delta respectively. To achieve the goals of the system, machine learning algorithms such as CNNs which is used for detection of spatial patterns of raw EEG data, and SVMs, which are used for feature handling and classification of small EEG data sets are used. This kind of dual-architecture guarantees the most correct and rapid affect recognition process. Subsequently, based on the categorical classification of the distinct emotional conditions, the system suggests the optimum kind of music on the basis of tempo, type, and emotional character. For example, stress is reciprocated by slow and smooth frequency music, while excitement is responded to by high pitch rate music. Real-time features also mean that the system is capable of responding to shifts in users’ emotions in real time, giving users an extremely personalized experience. Property-based validation proves that the system could be useful in music therapy, mental health surveillance, and entertainment. Larger scope of feature extraction, expansion of the database, integration of feature extraction with new opportunities in the field of virtual reality, games, and educational programs should become goals for future research.