Quantitative evaluation of college music teaching pronunciation based on nonlinear feature extraction

Abstract The complexity and diversity of different music styles created by different composers make music evaluation a very difficult problem. To quantitatively evaluate the quality of music pronunciation, this work proposes a method for extracting nonlinear features of music signals. By selecting different styles of music signals, based on chaos theory, fractal theory, and Lyapunov exponent, the change in correlation dimension of different kinds of music signals is analyzed from a nonlinear perspective, so as to realize vocal music evaluation and training in music teaching. The results showed that the filter used in the study has a good noise-reduction effect. The power spectrum of the music signal was qualitatively analyzed, and the music signal was characterized by chaos. The difference between the maximum and minimum values of the Lyapunov exponent of the For Alice music signal was 0.244, which is larger than the other music signals. After the correlation dimension analysis, the correlation dimension of the Croatian Rhapsody music signal varied strongly. The average correlation dimension of the five music signals fluctuated above and below the corresponding average correlation dimension value at order 0, and the music signals had stable nonlinear features. The research method of the article is conducive to deepening the understanding of music signal patterns and has a certain promotion effect on music teaching.