ScholarWorks@경상국립대학교

초록

Ultrasonic sensors have a simple operating principle, are compact, and are cost-effective. Due to these advantages, they are used in various application fields. In this paper, to recognize various hand gestures, one ultrasonic transmitting sensor and four ultrasonic receiving sensors were arranged in a square formation. The ultrasonic waves generated by the transmitting sensor are reflected by hand gestures and uniformly received by the four receiving ultrasonic sensors. The received ultrasonic reflection signals are stored as time-series data, and the stored data is trained using a CNN(Convolutional Neural Network) model. A dataset of ten hand gestures was selected for the experiment, consisting of five simple gestures and five complex gestures. Each class contained 500 validation data samples and 50 test data samples. The test results achieved an accuracy of 96%. © 2025 IEEE.

키워드