Playing the Chinese flute requires accurate coordination of blowing technique and fingering. This complex performance demands fine control over airflow velocity, embouchure angle, and the control of six finger mechanical to achieve accurate pitch and tonal quality throughout the musical performance. The robot features an artificial embouchure that simulation human lip formation, coupled with a two-way bellows system that alternately supplies air to apressure- regulating-bellows. A microcontroller-integrated system controls six mechanical fingers for hole covering and releasing during performance. To address the issue of air leakage at the end of notes, an air gate has been designed, which can also simulate tonguing techniques through its opening and closing movements, enabling more precise articulation of flute sounds. For sheet music recognition, the system collects musical score samples divided into three sets: staff lines, musical notes, and rhythm patterns. These are processed through a Multi-Task Learning (MTL) deep learning sturcture to develop a recognition model capable of identifying notes within three lines above and below the staff, as well as note values ranging from whole notes to sixteenth notes, including dotted notes. Through testing has demonstrated that when the musical score falls within the appropriate range, the system can achieve complete recognition, successfully convert the notation into note data, and transmit this data to the flute-playing robot for performance execution.