The Classification and Judgment of Abnormal Problems in Music Song Interpretation Based on Deep Learning

In this paper, we propose a deep learning-based method for classifying abnormal judgments of music song interpretation problems, using computer to record the singer’s voice and then analyze judge it with trained model, pointing out main problems that exist in singing process, which is practical significance self-learning related teaching guidance. We collected more than 300 singers’ audio, data was calibrated classified by professional researchers field music, mainly divided into seven categories: vocal pallor, inability keep up rhythm, running tune, lack breath, unclear spitting, narrow range (the highs can’t go lows come down), true false problems. The audio feature extraction process applies Short Time Fourier Transform (STFT), Mel Frequency Cepstrum Coefficient (MFCC), spectral prime extract features randomly selected part used as training validation dataset. This dataset residual neural network efficient network. experimental results our model show accuracy about 90.1%, achieves good results.