Discrete Channel Estimation by Integer Passing in Information Bottleneck Graphs

The aim of this work is to develop low complexity receivers with close to optimum performance that circumvent high precision signal processing in the hardware. In particular, we consider receivers that perform all signal processing only on output indices of a quantizer. For such completely integer based receivers this paper introduces a new discrete channel estimation based on integer passing in an Information Bottleneck Graph. Its design principle is the maximization of relevant information with the Information Bottleneck method. The proposed discrete channel estimation reveals new insights into the theory of integer passing in Information Bottleneck Graphs and links integer passing in Information Bottleneck Graphs to the passing of real valued functions in the sum-product algorithm. The performance of the discrete channel estimation is investigated in an earlier introduced discrete receiver chain for low density parity check encoded transmission over flat Rayleigh fading channels.