广义相关时延估计算法的自适应实现形式

Adaptive Realizing Form of Generalized Correlation Algorithm for Time Delay Estimation

认为广义相关时延估计算法的缺陷是依赖对输入信号及噪声先验知识 ,特别是对其功率谱的了解 ,而在实际应用中 ,往往缺乏这种先验知识 ,只能以估计值来代替 ,从而影响时延估计精度 .自适应滤波器一般只需要很少或根本不需要任何关于输入信号和噪声的先验知识且性能良好 ,若用自适应滤波器替代广义相关法中的预滤波器就可放松或免除对输入信号和附加噪声的条件限制 ,提高时延估计的精度 ,广义相关时延估计算法时域中的预滤波器就是其频域中的权函数 .从理论上证明了 ,任何一个广义相关权函数 ,都能够变换成一个或几个 Roth处理器的组合形式 .实际上 ,Roth处理器就是自适应滤波器实现的、能对原始带噪信号做最优化估计的维纳滤波器 ,从而实现了自适应广义相关时延估计算法 .

The generalized correlation algorithm for time delay estimation is based on the knowledge of input signals and noise, especially of their power spectrums. But in practice, the data are sometimes lacking. Generally, the data are replaced by their estimated values. So the precision of time delay estimation is decreased. An adaptive filter only needs a few or doesn't need any data of input signals and noise, and it has good performance. If the adaptive filter is substituted for the prefilter in generalized correlation algorithm, the restricted conditions of input signals and noise can be relaxed or released, and the precision of time delay estimation can be improved. In generalized correlation algorithm, the prefilter in time domain is the weighted function in frequency domain. The authors proved that a weighted function of generalized correlation can be transformed into a Roth processor or the combined form of several Roth processors in theory. A Roth processor is a Wiener filter which is realized by an adaptive filter, and it can finish optimum estimation of the original signals with noise. It realizes the adaptive generalized correlation algorithm for time delay estimation.