Research On Demodulation Method Based On K-means Clustering For Mobile Optical Camera Communication

With the popularity of internet of everything(Io E)and the intelligent mobile devices,optical camera communication(OCC)using image sensor have attracted more and more attention in the academic circles.The OCC system uses the light emitting diode(LED)to send signals,and the complementary metal oxide semiconductor(CMOS)camera to receive signals.There are some technical challenges for the demodulation of the received signals in the OCC system based on the rolling shutter effect,such as the fluctuation of grayscale values,“bloom effect”and so on.In addition,in order to facilitate the application of OCC systems in the real word,the scenarios enabling user's movement need to be considered.In this paper,considering the mobile OCC system,the problems of the existing demodulation schemes at the receiver end are analyzed in detail.The corresponding schemes are also proposed to improve the system performance.The main efforts are as follows:1.Two demodulation schemes based on convolutional neural network(CNN)and k-means clustering are proposed to alleviate the stripes distortion caused by the movement.The experiments are demonstrated in mobile OCC system,and the bit error rate of the two schemes using k-means and CNN are compared.The experimental platform is built in proportion to the actual of 1:10.Experimental results show that the bit error rate(BER)performance of the k-means based demodulation scheme is slightly better than that of the CNN based demodulation scheme,and the overall BER of k-means is lower than the CNN scheme at the lateral moving speed.When the moving speed is 60 cm/s,the BERs of k-means and CNN based demodulation methods are1.54×10-5 and 3.04×10-5,respectively.However,in the vertical moving scene,the BER of k-means based demodulation is higher than that of CNN based demodulation scheme.When the vertical moving speed is 60 cm/s,the BERs of k-means and CNN based demodulation methods are 6.02×10-4 and 3.94×10-4,respectively.2.To against the influences of signals overlap interference from multi-LEDs and the movement of the receiver,a demodulation scheme based on multi-column matrix selection and k-means clustering is proposed.The proposed scheme can avoid the signals interference,while alleviating the stripe distortion caused by movement.In addition,the effects of different moving speeds and different distances between the two LEDs on the performance of the proposed scheme are analyzed.The experimental platform is built in proportion to the actual of 1:10.The results show that compared with the traditional scheme,the proposed demodulation scheme significantly improve the communication performance and robustness of the mobile OCC system.At the distance of 10 cm between two LEDs and the moving speed of 20 cm/s,the proposed scheme can achieved bit error rate of 8.74×10^-6,which is well below the 20%forward error correction threshold(3.8×10^-3).As the moving speed increases,the distortion of the stripe widths increases and the performance of the proposed scheme decreases.Thus,the proposed scheme is more available in indoor scenario with low moving speeds.