一种“附近”空间关系增强的多源融合室内定位方法

A“near”relation enhanced multi-sourced data fusion indoor positioning method

针对传统室内定位模式单一,结合室内位置描述中常用的“附近”空间关系,融合多传感器数据,本文提出一种“附近”空间关系增强的多源融合语音交互室内定位方法。首先,研究“附近”空间关系特征,针对室内环境,建立基于“窃取面积”和最短距离的“附近”空间关系的概率密度函数;其次,采集每个参考节点的指纹信息及节点间的距离和运动信息,基于隐马尔可夫模型对室内位置描述定位过程建模,通过维比特算法预测用户位置;最终,通过实际场景对本方法验证,本文提出的方法平均定位精度在1.88 m,80%的情况下定位精度可以达到2.12 m。

Aiming at the problem of the single traditional indoor positioning mode,a“near”relation in locality description enhanced multi-sourced data fusion voice interaction method for indoor positioning is proposed.Firstly,the characteristics of“near”spatial relationship are studied.The probability membership function of“near”spatial relationship is established based on“stolen area”and the shortest distance for indoor environment.Secondly,the fingerprint information of each reference point,the distance and motion information between reference points are collected.The process of indoor locality description is modeled based on the hidden Markov model,and the user location is predicted by the Viterbit algorithm.Finally,the experiment show that the average positioning accuracy of the proposed method is 1.88 m,and the positioning accuracy can reach 2.12 m within 80%.