Autoregressive parameter estimation from noisy data

A least-squares based method for noisy autoregressive signals has been deve loped recently, which needs to neither prefilter noisy data nor perform par ameter extraction. In this brief, a more computationally efficient procedur e for estimating the measurement noise variance is developed, and then an e fficient implementation of the method is presented. It is shown that this b etter way of implementation can considerably reduce the computational requi rement of the least-squares based method without any performance degradatio n. Computer simulations that support the theoretical findings are given.