The life of opera works lies in the performance, and the performers on the opera stage can give new artistic expression to the artistic performance of the opera works through the expression of effective vocal skills and the integration of them with AI technology. On the basis of combing the vocal singing skills on the opera stage, the article discusses the relevant training methods of vocal singing skills. Then the vocal singing voice data is collected with 10 different types of opera works, and the signal preprocessing is carried out through the methods of preemphasis, frame-splitting, endpoint detection, etc., and then MFCC is used to extract the emotional features of the vocal singing voice. As the input of CLDNN network, different emotional features are fused with dimensionality reduction and weighting through linear layer, and SAGAN model is introduced to mine the temporal relationship in emotional features, and then realize the recognition and classification of emotional features of vocal singing speech. The results show that the CLDNN-ASGAN model’s classification accuracy of vocal singing emotional features reaches 88.75%, and the average score for different types of opera vocal singing speech reaches 91.99, with a difference of only 0.18 points from the score of professional listeners. Utilizing AI technology to assist vocal performers in clarifying the emotional performance of their works during the performance process can promote the enhancement of the artistic expression of opera works.
