Research Area:  Machine Learning

Abstract:

With the development of industrial information technology in recent years, gesture control has attracted wide attention from scholars. Various gesture control methods have emerged, such as visual control, wearable device control, magnetic field feature extraction control. Based on one of the visual gesture control methods, this paper proposes a visual gesture control method applied to music box control by combining YOLOv4 object detection network. We design seven main gestures, reconstruct gesture datasets, and retrain the YOLOv4 object detection network by the means of the self-built datasets, further build a music box gesture control system. In this paper, we obtain the recognition accuracy of 97.8% for the object detection network in the gesture control system after a series of experiments, and recruit eight volunteers to conduct experimental tests on the self-built gesture-controlled music box system, mainly to quantify the time of executing a single command, attention concentration, etc. The results show that compared with the traditional control method, the visual gesture control method ensures the accuracy while has a faster response speed and takes up less of the user-s attention.

Keywords:  
Gesture recognition
YOLOV4
Industrial
Feature extraction control
Object detection

Author(s) Name:   Zhuowen Zheng

Publisher name:  IOP Publishing

DOI:  10.1088/1742-6596/2196/1/012026

Volume Information:  

Paper Link:   https://iopscience.iop.org/article/10.1088/1742-6596/2196/1/012026/meta