The TOA (time of arrival) estimation algorithms based on match-filtering detection for UWB (ultra wideband) wireless sensor networks are extensively studied in this paper. Based on the analysis of the drawbacks of the algorithms in the literature, a three-step algorithm is proposed: first, determine the search region for DP (direct path) detection; then, a rough detection of DP is made by threshold comparison; and last, the precise location of DP, i.e., the center of the arriving pulse, is obtained by a refined search process. The threshold factor used to calculate the threshold in the second step is set dynamically by using a model in terms of the kurtosis of the match-filtering output. The model is well independent of the channel model, and its effectiveness is proved through the comparison of the resulted performance with that of using fixed threshold factor. By comparing the performance of this algorithm with that of other algorithms, it can be observed that the proposed three-step algorithm has achieved a good trade-off between computational efficiency and estimation accuracy, thus more appropriate for current applications. In addition, the reliability of TOA estimation result is discussed through statistical analysis. Two levels of reliability are defined with regard to the corresponding kurtosis of the TOA estimation, and the probability density function for TOA estimation errors of each level is modeled. Properly incorporating the reliability information into the positioning algorithm will definitely improve the final location estimation accuracy.