Multi-Detector Deep Neural Network for High Accuracy Wi-Fi Fingerprint Positioning

Abstract

A Deep Neural Network (DNN)-based positioning algorithm with multi-detector architecture is proposed for high accuracy Wi-Fi fingerprint positioning. Our DNN-based approach fuses the scalability of classifiers and the precision of regressors. Moreover, a pre-processing pipeline of signal readings is added for characteristic grouping and intra-sample normalization to improve the robustness. The algorithm was trained and tested on a robotically surveyed indoor fingerprint dataset including 349 reference points and 191 effective Wi-Fi access points in a 30 m × 12 m area. As a result, our algorithm is capable of positioning with 1.08 m mean distance error in a leave-10%-out test, performing nearly three times as good as the referenced WKNN baseline.