Modified Kalman Algorithm For Fast Channel Parameter Estimation Using Uncertain Data

Simultaneous parameter estimationldetection in a mobile communication system, for rapidly changing channels, can be peMormed by interconnecting a Kalman type recursive estimator with a Viterbi decoder. This combination works properly provided that (i) a training sequence is sent each time when the parameters change, or, (ii) the error rate of the detector is very small. However, in practice, the channel parameters change randomly and continuously, and the error rate may be high, therefore this classical scheme may easily diverge. The standard Kalman type optimum recursive estimation equations can be reoptimized to include decision errors and maintain stability of such a system for error rates higher than 10%. For this purpose an Erasure Declaring Viterbi Algorithm which provides side information about the reliability of the data may be used.