Early detection of target manoeuvres under a specific false alarm rate

Two improvements in fading memory average for manoeuvre detection are given. First, according to the characteristics of fading memory average, the exact detection threshold is derived by a Markov chain approach. In conventional algorithms, which ignore the dynamic and dependent properties of fading memory average, only approximate threshold values are computed for a given false alarm rate. Secondly, the univariate fading memory average is further extended to the multivariate form to increase the speed of detection. The multivariate Markov chain approach, which is an extension of the Markov chain approach, can provide exact thresholds for the multivariate fading memory average. Monte Carlo simulation results verify not only the accuracy of the thresholds through both the univariate and multivariate Markov chain approaches, but also the superior detection speed of the proposed multivariate fading memory average. Under the same false alarm rates, the mean time to detection in the multivariate algorithm is 26% less than that in the univariate algorithm.