两级参考点匹配位置指纹声源定位方法

提出一种两级参考点（RPs）匹配方法来减少位置指纹声源定位（SSL）过程中临近参考点搜索的计算量. 离线采样阶段：通过K均值聚类算法将数据库划分为一定数目的子库，并采用一种距离检测方法对离群点进行剔除. 在线定位阶段：通过第一级临近子库匹配完成对参考点搜索范围的缩减；在临近子库内进行第二级参考点匹配得到临近参考点；完成声源目标（TP）定位. 实验结果表明，采用两级参考点匹配算法可以在保证定位精度的前提下有效提高位置指纹声源定位方法的定位效率.

Fingerprint-based sound source localization method using two-stage reference points matching

A two-stage matching method was proposed for computation reduction of adjacent reference points (RPs) searching in fingerprint-based sound source localization (SSL). In offline sampling phase, the K-means clustering method was adopted to divide the database into a certain number of sub-databases and the outliers were eliminated by the distance-based detection method. In online positioning phase, searching space was compressed by the first stage sub-database matching; then adjacent RPs were obtained through the second stage RPs matching in the adjacent sub-database; the auditory target point (TP) location estimation was accomplished. The experimental results show that the two-stage RPs matching algorithm can effectively improve the positioning efficiency of fingerprint-based sound source localization on the premise of ensuring the positioning accuracy.