Artificial neural network-based threshold detection for OOK-VLC Systems

This paper presents new detection threshold methods to improve the On-Off Keying (OOK) receiver scheme. In the paper, three definitions are discussed considering low-mobility, fast-mobility, and non-mobility scenarios: Integration Method (IM), Artificial Neural Network (ANN) Method-1 and ANN Method-2. For non-mobility scenario, we use IM which has interconnected structure since the receiver uses a test signal to determine the threshold value. In low-mobility case, the ANN Method-1 is very successful compared to ideal system which is completely knows the threshold value. According to simulation results, ANN Method-1 significantly improves Bit Error Rate (BER) performance at a 2.25 m distance. Therefore, the communication distance can be increased from 2.25 m to 2.52 m at a BER of 10(-3). Moreover, we think that the received optical power can suddenly change depend to dimming level for simulation and practical environments. The ANN Method-1 cannot detect the threshold value when the percent deviation of threshold level is higher than 100%. In order to solve this problem, ANN Method-2 is proposed in the paper. From simulation and practical results, it is shown that ANN method-2 is successfully detects the threshold value from received signal for 200% or more deviation. The proposed methods are designed on Field Programmable Gate Arrays (FPGA) board to observe real-time results. From simulation and practical results, it is shown that BER performance of ANN Method-2 is very close to BER performance of ideal receiver scheme.