With the continuous progress of art education and artificial intelligence technology, traditional music teaching models are facing transformation. This article aims to construct an art education and teaching system based on artificial intelligence, especially for teaching music sound recognition. Through in-depth research, we have designed a music sound recognition system that uses Mel frequency cepstral coefficient (MFCC) for feature parameter extraction, and combines BP neural network algorithm to construct a music sound learning model. The main purpose is to improve the efficiency and accuracy of music teaching through artificial intelligence technology. The main challenge we face in this process is how to effectively extract the features of music sounds and accurately identify different tones through algorithms. By using the MFCC algorithm, we have successfully solved this problem as it can effectively describe the time-frequency characteristics of music sound. Our proposed music sound learning model is based on a BP neural network, which trains the network to learn the mapping relationship between music sound and pitch. The experiment used piano sound as an example to verify the accuracy and reliability of the system. The simulation experiments conducted in MATLAB environment show that our system can accurately recognize and extract the main frequency of music, and has higher performance compared to traditional methods.