The goal is to promote the healthy and stable development of music education in China. The time-frequency sequence topology in frequency domain can improve the effect of convolution operation. Therefore, this paper applies the above algorithms to classical music education, including the recognition of classical instruments, the feature extraction and recognition of classical music, and the quality evaluation of classical music education. The quality of the music quality evaluation system can be judged according to the correlation between the output results and the subjective evaluation. The higher the correlation, the better the music quality evaluation method. Through relevant experiments, it is proved that DTW score alignment and end-to-end are more successful in extracting the features of classical music, and more accurate in identifying classical instruments. The objective evaluation method of pronunciation teaching quality is more objective and accurate than P.563 music teaching quality evaluation.