ETRI-Knowledge Sharing Plaform

논문 검색
Type SCI
Year ~ Keyword


Conference Paper Punch Analysis with FFT and LSTM of Accelerometer and Gyroscope Data
Cited 1 time in scopus Download 8 time Share share facebook twitter linkedin kakaostory
Park Hyunho, Kwon Eun Jung, Byon Sungwon, Shin Won-Jae, Jung Eui Suk, Lee Yong Tae
Issue Date
International Conference on Information and Communication Technology Convergence (ICTC) 2020, pp.1353-1355
Project Code
20HR3200, Platform Development of Multi-log based Multi-Modal Data Convergence Analysis and Situational Response, Lee Yong Tae
Governments around the world are interested in public safety services for protecting the public from violent crimes. For the public safety services, a technology for detecting violent activities attracts attentions. This paper proposes an analysis method for analyzing punches by using fast Fourier transform (FFT) and long short-term memory (LSTM) of accelerometer and gyroscope data of wearable devices. The proposed method analyzed characteristics of accelerometer data in frequency domain by using FFT. The proposed method achieves 97.27% accuracy and 97.27% F1-score for classifying types of punches (e.g., left hook, left jab, right cross, and right hook) by using LSTM. The proposed method will contribute to detect violent activities and thus will help fast response to the violent crimes.
KSP Keywords
Accelerometer and gyroscope, Accelerometer data, Analysis method, F1-score, Fast fourier transform (fft), Long-short term memory(LSTM), Public safety, Violent crimes, Wearable device, fast response, frequency domain(FD)