中文摘要：

当环境信噪比低于某一阈值,由输入噪声引入的有偏维纳解使最小均方（LMS）自适应时延估计器的错误时延估计概率陡然增加。系统分析了LMS自适应算法有偏估计的噪声来源及有偏时延估计器的性能,并基于Treichler的γ-LMS算法提出了一种修正无偏自适应时延估计方法。该方法利用传统LMS自适应滤波器可以获得的信息估计输入噪声功率,然后根据γ-LMS算法的思想在迭代过程中引入一个修正因子,逐步去除输入噪声的影响,无需假设输入与输出噪声功率相等或功率比已知、有用信号应为白过程等限制条件。自适应时延估计的仿真实验与供水管道泄漏检测定位中的实际数据处理都验证了该方法的有效性。

英文摘要：

Below a certain signal-noise-ratio （SNR）, the biased Wiener solution induced by the input noise will result in that the probability of wrong estimation increases rapidly in least-mean-square （LMS） adaptive time delay estimators. This paper systematically analyzes where the bias comes from and the performance of the biased estimator. A modified bias-free algorithm is developed based on Treichler＇s γ-LMS algorithm. The input noise power is first obtained with available information from traditional LMS algorithm, then by introducing a γ factor related to the input noise power in adaptation process the input noise effects can be iteratively eliminated. The proposed method needn＇t assume that the input and output noise powers are the same or their ratio is known, or the signals are all white processes. Simulations and real data applications in water pipe leak detection and location are provided to validate its effectiveness in an adaptive time delay estimation system.