Laser & Optoelectronics Progress, Volume. 58, Issue 24, 2428004(2021)
Research on Adversarial Examples in Human Physical Rehabilitation Exercises Based on GPREGAN Framework
Figures & Tables(8)
Fig. 3. Recognition rate of depth model. (a) CNN recognition rate; (b) LSTM recognition rate
Fig. 4. Recognition rate of depth model for adversarial examples. (a) CNN recognition rate; (b) LSTM recognition rate
Fig. 5. Distance between adversarial example and original sample. (a) CNN; (b) LSTM
Table 1. Influence of SNR on depth model
View tableTable 1. Influence of SNR on depth model
SNR -8 dB -6 dB -4 dB -2 dB 0 2 dB 4 dB 6 dB 8 dB CNN 0.7625 0.8175 0.9100 0.9300 0.9600 0.9725 0.9775 0.9850 0.9875 LSTM 0.2600 0.3625 0.4475 0.5375 0.7275 0.7525 0.8100 0.8700 0.8925
Table 2. Mean square error between original sample and adversarial example
View tableTable 2. Mean square error between original sample and adversarial example
Group No. 1 2 3 4 5 6 7 8 CNN 0.0163 0.0154 0.0147 0.0139 0.0132 0.0127 0.0122 0.0119 LSTM 0.0008 0.0168 0.0108 0.0127 0.0049 0.0134 0.0047 0.0175
Tools
Get Citation
Copy Citation Text
Kangjie Zheng, Shan Jin, ChengWei Zhang. Research on Adversarial Examples in Human Physical Rehabilitation Exercises Based on GPREGAN Framework[J]. Laser & Optoelectronics Progress, 2021, 58(24): 2428004
Paper Information
Category: Remote Sensing and Sensors
Received: Jan. 5, 2021
Accepted: Mar. 2, 2021
Published Online: Dec. 3, 2021
The Author Email: Zhang ChengWei (chenvy@dlmu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20