Use of Kalman filtering in data detection in optical communication systems with multiplicative noise

It has been shown that in optical communication systems with detectors that have a multiplicative noise component, the bit error rate (BER) can be improved by orders of magnitude through the use of a thresholding scheme based on a likelihood ratio test (LRT) instead of a matched filter. An LRT detection scheme of this type has been derived previously for optically encoded Manchester data; however, the development of the threshold required that the means and variances of the bit levels were known. In free-space communication systems, atmospheric turbulence may cause large variations in the optical transmission and subsequently in the means and variances of the bit levels. A dual Kalman filter-based method has been developed which uses an adaptive LRT threshold for bit detection while tracking the mean and variance of the high and low bits. The development of the Kalman filter algorithm is discussed and the results of simulations comparing the performance of the adaptive algorithm to a matched filter and to a non-adaptive LRT detection scheme are presented.