基于摄像头的多光源多输入多输出(MIMO)通信系统能够提高光源的利用率和传输带宽, 但多光源之间的干扰和环境光的干扰为可见光的颜色判断带来了极大的困难。针对这种情况, 提出了基于自联想神经网络的可见光MIMO颜色检测算法。设计了基于自联想神经网络的主成分提取方法, 在训练速度和模型准确率方面取得平衡; 提出了神经网络学习光源颜色的具体训练方法,基于欧氏距离决定光源的实际颜色符号。在真实的实验数据集的基础上完成了验证实验, 结果显示该算法实现了较好的收敛速度和较高的检测准确率。

It is possible to improve both of the utilization ratio of light source and the transmission bandwidth of multiple light source multiple input and multiple output by camera, but the multiple light source interference and environment light interference bring great difficulties to color decision of visible light communication. In view of this, a color detection algorithm of visible light MIMO communication system based on the auto-associative neural network is proposed. First of all, a principal component extraction method based on the neural networks is designed, a good balance between the training speed and the accuracy of the neural networks is realized; Secondly, the detailed training procedures of light source color learning by neural networks is presented; Finally, it makes a decision for the real color symbol based on Euclidean distance. Several validation experiments based on the real experimental datasets are completed, the results show that the proposed algorithm realizes fast convergence speed and good detection ratio.