| Time delay estimation is an active research field in signal processing, and it has important meaning not only at theory but also on application. Time delay estimation brings new require on random signal processing, modern spectrum estimation, time sequence analysis, adaptive signal processing, some interrelated technology and Fourier transform and some progress about these subjects and technology has been obtained. With the development Time delay estimation has been made great use on military field such as radar and sonar. It also made great use on oil exploring, detection of underground piping leakage, malfunction diagnoses of oscillation, hydrophilic field, and seismology and biology astrology and so on. Great effort has been made about time delay estimation. Based on the consulting and analysis of abundant reference materials, main time delay estimation methods are summarized. The paper introduces many methods of time delay estimation and gives the relation between these methods. The application of LMSTDE is analyzed including the based principle,arithmetic structure and system performance of LMSTDE and so on. Through the computer simulation, we can observe that the method of LMSTDE applies to the condition that the signal and the noise power is invariable. When signal and noise power is variable, the result is degradation in system performance, and we also can see that the ability of tracking has been weakened. These are shown on the simulation experiment .In order to solve this problem; we should find an advanced method. Previous work in LMSTDE has only considered the situation where signal and noise power are constant. That means that any change of the filter parameters is due solely to variations in time delay .however, through analyzing the LMSTDE method, we can see that the weight coefficient can been shown with the next formula: W*=αSNR×(1+SNR)-1[sin c(-P-D) sin c(-P+1-D)...sin c(P-D)]TIn actual circumstances, however, when the SNR is low and time varying due to changes in signal power and noise power, the adaptive filter will keep track of both the delay and SNR. Because of coupled adaptation between these quantities, the time constant will no longer represent the actual learning behavior ofD?, which is not a desirable feature. In addition, a slowdown in convergence of delay parameter is observed. Furthermore, a small SNR will make the filter weights to have small magnitudes .producing a large delay variance. As a result, the system performance is greatly deteriorated. If we still use the LMSTDE method, simulation results show that the system attenuation is serious and the tracking ability is also weakened. The single adaptive filter W (z)in the conventional model is separated into two adaptive subunits A( z)and g connected in cascade,(shown on the following figure) where A( z) is to produce the appropriate time shift while the adaptive gain g provide proper scaling to the filter output one for tracking the delay and the other for adapting the SNR. This arrangement leads to a decoupled adaptation of SNR and delay to fit the timing change of delay and SNR. Through comparison and analysis with the LMSTDE, simulation results show that the method proposed in the paper has better performance. These performances include delay error, system tracking ability and speed of convergence.
|
PDF Full Text Request