Stochastic averaging analysis of a steepest-descent-type adaptive time- delay estimation algorithm (Q1842528)

Consider the following signal model; \(x_{1k} = s_ k + n_{1k}\); \(x_{2k} = s_{k + \theta_ k} + n_{2k}\), where \(s_ k\) is the noise-free signal, \(n_{1k}\) and \(n_{2k}\) are noises, and \(\theta_ k\) is the time-delay to be estimated. The authors present a stochastic averaging analysis of the adaptive steepest-descent time-delay estimation algorithm, which extends the averaging results of \textit{H. J. Kushner} [cf. Approximation and weak convergence methods for random processes, with applications to stochastic systems theory, MIT Press (1984; Zbl 0551.60056)] to cover the state-dependent noise case. It is proved that as the adaptive gain \(\mu \to 0\), the adaptive time-delay estimator converges weakly to the unique solution of an ordinary differential equation and that under an ideal situation the solution of this differential equation converges exponentially to the true delay. Practical implementation of the algorithm, a modified first-order approximation unbiased algorithm, second-order analysis and computer simulation are also included to support the analysis.