Deep Neural Network based Music Source Conducting System
Myungin Lee · University of Michigan Library Repository · 2018
In this paper, we propose a system that can interact with music source In this paper, we propose a system that can interact with music source using a microelectromechanical systems (MEMS) based controller which is inspired by the conducting activity. Conducting is one of the most exquisitely developed connections between music and gestural activity. The system gives inactive and intuitive musical experience to the user with the existing music source using a smartphone with MEMS sensors. By using a deep neural network (DNN), the algorithm derives the temporal, amplitude, and filter information at the same time from the input data. While conventional studies on the analysis of conducting gesture have a limitation on modeling complex model, the proposed system classifies various messages with high accuracy. The real-time demonstration of the system is provided through MATLAB.