基於壓縮感知及通道響應於室內定位技術之設計與分析

Abstract

Since the accuracy of indoor location estimation environment with weak GPS signals is low, indoor localization topics have attracted more and more attention in recent years.In this paper, a novel compressive sensing based location estimation using channel impulse response measurements (CS-CIR) is proposed as well as fingerprinting algorithm. In off-line stage, reference points (RPs) will receive signals from access points (APs) and send them to the network center (NC) for create the database. During the on-line stage, the user equipment (UE) measures CIR information from APs and then quantizes the measured CIR by $B$-bit quantizer for reducing transmission overhead after doing compression according to compressive sensing. The UE sends the quantized signal to NC as the feedback, and NC derives the UE's position by optimization methods or the greedy method. Moreover, we derive a orthogonal basis pursuit (OMP)-based lower bound of correct identification probability under the assumption that the UE is on the top of RP which is a 1-sparse signal having an exploit relation to quantization error and the channel Gaussian noise effect. This bound guarantees that the OMP-based correct probability on $1$-sparse signal can perfectly identify the signal in high-SNR regime under the assumption. Simulation results validate that the CS-CIR outperforms the $K$-nearest neighbour method using CIR measurements and conventional received signal strength based methods. In addition, the OMP-based lower bound of correct identification probability is effective to our proposed method.

由於 GPS (Global Positioning System)訊號應用在室內定位的低精確度,室內定位研究議題近年來已經逐漸受到關注。在這篇論文中，我們提出一種新穎的壓縮感知 (compressive sensing)結合通道響應 (channel impulse response)與指紋辨識法 (fingerprinting) 於位置估計。在離線階段時，每個參考點 (reference point (RP)會收集其他無線接取器 (access point (AP)所傳發射的訊號，並回傳給網路中心建立資料庫。在上線階段時，使用者會接收到所有無線接取器的訊號並根據壓縮感知理論將此訊號壓縮以降低訊務上的負擔，並且用B位元量化器做量化。接著將此量化訊號回傳網路中心做為回饋，而網路中心藉由最佳化方法及貪婪法求得使用者的位置資訊,並藉由壓縮感知可利用解最佳化方法或貪婪法將使用者的座標求出。 此外，假設使用者出現在參考點上並其位置可描述為一單位稀疏訊號，我們推導出在此假設下基於正交匹配追蹤法之正確機率下界，此下界與量化誤差及雜訊有關，並且保證在高訊雜比下，利用正交匹配追蹤法可以完美的還原原始訊號。 我們利用模擬結果來驗證本篇論文提出的方法,使用通道響應之壓縮感知法(CS-CIR)之效能會比使用通道響應之最近鄰居法 (KNN-CIR)及其他傳統使用訊號強度的方法來的佳。除此之外，此基於正交匹配追蹤法的正確機率下界對於本篇論文所提出的方法是有效的。