Fig. 5: Posture recognition.
From: Multilayer nonlinear diffraction neural networks with programmable and fast ReLU activation function
a Schematic of MN-DNN for posture recognition, where the scattered EM fields from the test person are processed by MN-DNN. b Equivalent diagram of MN-DNN. c Experimental setup for posture recognition. d Schematic of the detection setup on the output plane, consisting of eight zones each equipped with an antenna and detector that convert EM energy into DC voltages (V1–V8). These voltages are processed by ADC of MCU, controlling the LED colors based on the maximum voltage channel, and transmitted to a computer for real-time display. e Experimental results from 50 repetitions of eight postures, showing the accurate classification as confirmed by voltage readings across the eight regions.
