Indoor Localization Based on Semi-tensor Product Compression Sensing

The sparsity of the localization problem makes the Compression Sensing (CS) theory suitable for indoor localization in Wireless Local Area Network (WLAN). However, when the target environment has a large number of Access Points (APs), online measurement takes a lot of time and increases the storage space of the AP selection matrix. To address this drawback, a Semi-tensor Product Compression Sensing (STP-CS) model is proposed, which reduces the time required to construct the AP selection matrix by constructing a low-order matrix, and enhances the real-time performance of online localization simultaneously. Experimental results show that, compared with the traditional methods, the proposed method has a great improvement in real-time performance and low overhead while ensuring localization accuracy.

I. Introduction

With the development of Internet of Things (IoT) technology, Location-Based Service (LBS)[1] has been used in various fields, in which positioning technology is one of the most important key technologies. Due to the pervasiveness of Wireless Local Area Network (WLAN), indoor localization systems based on WLAN have become one of the most common methods for indoor environment. Generally, there are two types of indoor localization techniques, one is based on signal spatial similarity and the other is based on sparse recovery technique, where the Compressive Sensing (CS) is mostly utilized. Yan[2] designed a localization scheme to enforce the Restricted Isometry Property (RIP) and maximize the signal-to-noise ratio (SNR) by introducing optimization of the measurement matrix. Khan[3] proposed a method to estimates the location of unknown/target nodes using Bayesian CS. However, when the target environment deploys a large number of Access Points (APs), the AP selection matrix constructed by the traditional CS methods[4] will increase computational complexity and storage space. Hence, this paper proposes a novel Semi-tensor Product Compression Sensing (STP-CS) model, which satisfies the two criteria of RIP and Mutual Incoherence, and can obtain an equal localization performance as traditional CS but reduces the storage space and computational complexity by constructing a low-order random measurement matrix, and realizes real-time and accurate location estimation.