Indoor Positioning Model Based on Principal Component Extraction of Support Vector Regression

The use of the WLAN Received Signal Strength Indicator (RSSI) for indoor positioning has gained significant attention due to its advantages such as cost-effectiveness, low hardware power consumption, and ease of implementation. However, the presence of multipath and shadow effects often hinders traditional positioning algorithms from achieving satisfactory results. To address this challenge, the paper proposes a method that utilizes a feature data dimensionality reduction algorithm to extract features from the raw data, thereby improving positioning accuracy. This method utilizes relative errors as characteristics of the fingerprint database to establish a regression map between RSSI values and positioning coordinates using Support Vector Regression (SVR). The algorithm's positioning performance is evaluated in simulation environments, and the study analyzes the factors that influence the algorithm's performance. Experimental results demonstrate that the Kernel Principal Component Analysis (KPCA)-SVR algorithm outperforms other methods, in terms of higher positioning accuracy and better robustness. This approach shows promising potential for enhancing indoor positioning accuracy and overcoming the limitations of traditional algorithms in challenging indoor environments.