Research On Time Delay Estimation Method Of Narrowband Signal

Time delay estimation has wide applications in many areas, among which locating has astrict requirement for estimation precision. However, the estimation precision in the narrowbandsignal environment is not so desirable as in the wideband signal environment, and the resolutionprobability is low when using the traditional multipath time delay estimation methods innarrowband environment. As to the above problem, this paper aims at studying the way toimprove time delay estimation precision both in gaussion and stable distributed noises, aswell as the method to exalt the resolution probability under multipath effect based on thenarrowband signal environment. Besides, parts of the algorithms are implemented in GPU. Themain content is as follows:1. The way to improve time delay estimation precision of narrowband signals in Gaussiannoises is studied. A second correlation method based on wavelet denoising is presented. Thismethod denoises signal with time delay information kept in the light of wavelet denoising theory,combined with the second correlation method to get the time delay estimation. The simulationresults show that the proposed method performs better than the traditional correlation methodand the second correlation method, thus improving the precision of time delay estimation innarrowband signal environment.2. The methods for improving time delay estimation precision of narrowband signals in stable distributed noises are discussed. A class of covariation time delay estimation methods anda class of fractional lower order methods based on preprocessing functions are proposed. Thetheoretical analysis indicates that the former method has a finite variance by preprocessing thereceived signals with any odd symmetry, monotone increasing and bounded functions, whichimproves the accuracy of time delay estimation by a large margin, making the apply of themethod in practice become possible. The preprocessing functions and their conditions are givenin the latter method. Through mathematical reasoning the conclusion is drawn that the varianceof the fractional lower-order covariance of received signals decreases and the time delayestimation is still unbiased by preprocessing them with corresponding functions, which enhancesthe detection precision of the peak of the fractional lower-order covariance, thus leading to theimprovement of time delay estimation precision. Two preprocessing functions are introduced inthe end with regard to the two kinds of methods, respectively. The simulation results indicate thatthe two proposed methods are effective when estimating time delay of narrowband signals in stable distributed noises as well as in Gaussian noises.3. The research is also focused on the time delay estimation method of narrowband signals under multipath effect with high resolution probability. A sparse signal estimation method basedon prominent eigenvector is proposed. This method takes the advantages of the sparse signalsinto multipath time delay estimation and the simulation results show that the proposed method isstill feasible when samples are small and even when the delay between different paths are farsmaller than the resolution probability. What‘s more, the estimation precision reaches highenough.4. The covariance covariantion and the fractional lower order time delay estimationmethods based on preprocessing functions are implemented on GPU. Through analyzing theparallelism of these two methods, the time used on GPU and CPU are compared in detail. Theresults validate the procedures on GPU and show that the running time is decreased, thus provingthe methods implemented on GPU can be real-time.