Summary: We study the data compression technique which is used in radar measure systems to identify and track moving targets precisely. First, linear predication coefficients \(a_i\) which are independent of statistical feature of signal source are evaluated and a decision theorem is proposed. Secondly, two new methods based on predication technique are introduced to real-time lossless compression for data with significant correlation. Coding speed, time-delay, error bit propagation, self-synchronizing and compression rate are also discussed. A simulation test shows that the new algorithms are superior to existing algorithms.