The system comprises an input device for collecting sound and sound information or extracting sound information from a music sample; a pre-processor for pre-processing the informational collection to generate an input information test set for a characterization model, wherein the pre-processor utilizes fine-grained division and different techniques to preprocess the example informational collection; a central processor for combining sound feeling data and further developing arrangement speed, such that review makes fine-grained division for genuine music informational collection and results the inclination results by casting a ballot direction, which is configured to promote precision of music feeling grouping; a vocal division device for dividing vocal of the complicated structure of genuine music sound, and voice and foundation sound are incorporated together; and a reviewing device for reviewing the vocal detachment of music and reviewing the grouping impact of vocal and foundation sound individually, which incredibly builds the convergence of sound elements.