Transfer Regression Model for Indoor 3D Location Estimation

Junfa Liu, Yiqiang Chen, Yadong Zhang

Abstract: Wi-Fi based indoor 3D localization is becoming increasingly prevalent in today’s pervasive computing applications. However, traditional methods can not provide accurate predicting result with sparse training data. This paper presented an approach of indoor mobile 3D location estimation based on TRM (Transfer Regression Model). TRM can reuse well the collected data from the other floor of the building, and transfer knowledge from the large amount of dataset to the sparse dataset. TRM also import large amount of unlabeled training data which contributes to reflect the manifold feature of wireless signals and is helpful to improve the predicting accuracy. The experimental results show that by TRM, we can achieve higher accuracy with sparse training dataset compared to the regression model without knowledge transfer.