A Received Signal Strength Based Indoor Localization Algorithm Using ELM Technique and Ridge Regression

In this paper, a received signal strength (RSS) based localization algorithm by extreme learning machine (ELM) technique is proposed. In the offline phase, in order to cope with the environmental noise of the training data set, the improved ridge regression based ELM (IRR-ELM) algorithm is proposed to obtain more stable prediction which has better generalization ability, because the ridge parameter is obtained with the variance of the training error. In the online phase, the obtained prediction model is straightly used for position prediction. Since it can reduce the deviation in the off-line training phase, the proposed algorithm has more stable and accurate position estimation result in online phase. At last, the RSS measurements obtained from field test are used for performance evaluation. It shows that the localization performance of the proposed is better than that of the existing ELM based localization methods.