Abstract: To solve the problem of positioning accuracy deviation and poor robustness for indoor positioning based on WiFi fingerprints, causing by the heterogeneity of devices, a calibration-free indoor location method is proposed for across heterogeneous devices. The raw fingerprint database is standardized, combining the strongest AP classification and Procrustes Analysis. Training classification standardized fingerprints by using ELM（Extreme Learning Machine） method, a classified regression model for RSS（Received Signal Strength） and location is obtained, then it gets the positioning location. In the typical laboratory building, four heterogeneous types of mobile phones are used for experiments. The experimental results show that the proposed method improves the accuracy and stability of positioning, compared with traditional calibration free methods.

Key words: indoor location, heterogeneity of devices, strongest AP classification, standardized fingerprint, extreme learning machine

摘要： 针对基于WiFi指纹的室内定位中设备异构带来的定位精度偏移和鲁棒性差的问题，提出一种免校准的跨异构设备的室内定位方法。结合最强AP（Access Point，接入点）分类和普氏分析（Procrustes Analysis）对原始指纹库进行标准化处理，再经过极限学习机（Extreme Learning Machine，ELM）训练，建立分类的回归模型估计待定位节点的位置。在典型实验楼场景使用四种异构类型的手机进行实验，实验结果表明，与传统的免校准方法相比，该算法提高了定位的准确性和稳定性。

关键词: 室内定位, 设备异构性, 最强AP分类, 标准化指纹, 极限学习机