Machine-learning-assisted music conducting practice system

In this research, we introduce a machine-learning-assisted conducting practice system. By using the capabilities of machine learning in this system, conducting instructors are empowered to train a conducting gesture model to make the system capable of recognizing conducting movements. The system can capture the trajectory of conducting movements using mobile phones or three-axis sensors and play MIDI-based music, allowing trainees to practice movements using a readily available device easily. Moreover, our system divides the conducting gesture into several equal segments according to the meter of the music. This innovation can help the MIDI-based music more closely align with the speed of the user's conducting gestures, reducing the latency issues that otherwise often arise during virtual practice in other similar systems. Briefly, the system not only serves to facilitate trainees in practicing the movements but also enables novices to experience the artistic allure of conducting.