ANN-assisted Real-time Blind Signal Detection over Time-varying Doubly-stochastic Poisson Channel for Indoor Optical Wireless Communications

In this letter, we present a new real-time blind signal detection technique for indoor optical communications using a supervised learning framework in an artificial neural network (ANN). We model the optical channel as a time-varying doubly-stochastic Poisson process. To compare the performance of the proposed ANN network, we present an iterative method for joint channel estimation and symbol detection using expectation-maximization (EM) and Viterbi decoding and demonstrate that the bit error rate (BER) performance of the proposed technique is better than that of the Viterbi decoding. The proposed scheme is found to be resilient to indoor channel dynamics and can achieve a good BER performance over a wide range of channel variations.