Quantitative Analysis and AI Assessment of Piano Performance Techniques

Piano performance is a fusion of emotion and art, which needs to be expressed by adopting superior skills, but also to create the corresponding mood atmosphere to comprehensively show the infectious power of piano performance. This paper proposes the application of AI recognition of human bone and joint movements in piano performance, using BP neural network to realize the recognition of human bone and joint movements. The fingering action in piano performance is taken as a case study, and a bidirectional recurrent neural network is used to fit the relationship between pitch difference and fingering. A research paradigm for AI recognition of piano performance hand shape was also designed to map the discrimination of piano performance hand shape into an image recognition problem of hand gesture for fingering evaluation, driven by action data and knowledge of piano hand shape. In this paper, 28 short piano scores of Bach, 5 scores of Cherny 299 and 7 scores from the China Conservatory of Music’s Social Art Level Examination Grade 1-3, totaling 40 scores, are collected as the experimental dataset. In terms of algorithm performance, through experimental verification, the algorithm in the paper compared with the existing annotation model in the consistency rate and two new indicators to improve the effect is significant, the consistency rate can reach 69.7%, the percentage of incorrect fingering is reduced to 0%, and the rate of irrationality is reduced from 19.6% to 3.87%. In addition, this paper proposes a piano fingering evaluation method based on AI action recognition, which can objectively describe the advantages and disadvantages of piano fingering from the overall results, and has the effect of improving the performance skills.