Aiming at the currently used parameters are mainly static features, a dynamic feature calculation method based on information theory is proposed, and the initial value in key frame extraction is set according to its physical meaning, and the A algorithm is used to decompose the tensor, and then a tensor-based model is proposed. classifier. The experimental results show that the characteristics of the tensor model have a certain improvement effect on the problem of violent audio classification. National vocal music is the concentrated expression form of the Chinese national characteristic culture, showing many characteristics and potentials. Recognizing the strengths and weaknesses of each other's culture establishes a theoretical foundation, finds a new theoretical fulcrum for re-evaluating the value of different music cultures fairly and reasonably, and provides the necessary international quality for cultivating talents in the new century.