With the development of information technology, robot technology has made
great progress in various fields. These new technologies enable robots to be
used in industry, agriculture, education and other aspects. In this paper, we
propose a drum robot that can automatically complete music transcription in
real-time, which is based on AIoT and fog computing technology. Specifically,
this drum robot system consists of a cloud node for data storage, edge nodes
for real-time computing, and data-oriented execution application nodes. In
order to analyze drumming music and realize drum transcription, we further
propose a light-weight convolutional neural network model to classify drums,
which can be more effectively deployed in terminal devices for fast edge
calculations. The experimental results show that the proposed system can
achieve more competitive performance and enjoy a variety of smart applications
and services.

利用人工智能物联网和边缘计算技术，提出了一个鼓机器人系统，可以实时完成音乐转录并通过轻量级卷积神经网络模型进行分析，以实现更高效的终端设备上的快速边缘计算和更多智能应用和服务。