Extreme Learning Machine and AdaBoost-Based Localization Using CSI and RSSI

Abstract

This letter develops an extreme learning machine (ELM) and AdaBoost technique for indoor localization using channel state information (CSI) and received signal strength indicators (RSSI), aiming to (a) resolve an issue with ELM in which weight parameters are generated randomly that leads to significant performance variations, (b) to maximize localization performance given the same set of measurement data, and (c) reduce the data storage and computational needs of existing schemes. To this end, we first form the fingerprint training dataset that consists of both CSI and RSSI, and fully extract from it the statistical information such as quartile, kurtosis, skewness, entropy, etc., in addition to the commonly used channel mean and std. The random weight issue of ELM is solved by optimizing the AdaBoost algorithm to combine multiple weak ELM classifiers and generate a stronger classifier for offline training. We provide experimental results to assess the effectiveness of the proposed scheme in terms of localization accuracy and storage/complexity.