A music station is connected through a communication network to another music station, and pieces of music data expressing an exhibition performance on a automatic player piano and pieces of voice data expressing tutor's explanation are transmitted from the music station to the other music station through different communication channels; and a close-talking microphone and a bone conduction microphone are incorporated in a sound collector on the music station, and a vibration signal from the bone conduction microphone is examined to see whether or not the cord of tutor vibrates; when the answer is given affirmative, a voice signal from the close-talking microphone is relayed to a transmitter module so that the sound collector does not permit the transmitter module to transmit the voice signal expressing noises such as the tones; whereby the music performance system prevents the trainee from tones reproduced from a headphone.