A novel algorithm for enhancing accuracy of indoor position estimation

In this paper, we propose an algorithm called SVM-WKNN for precisely indoor positioning, which can be applied to intelligent robot and wireless sensor network to achieve great performance. The key issue of indoor positioning is how to use the instable wireless network and nonlinear wireless signal strengths to accurately locate the position of a person or object. However, the traditional linear methods, such as calculating the signal distance, are failed in dealing with this problem. So we introduced the SVM which shows a great advantage in solving the nonlinear problem to achieve the precisely indoor positioning. Because the stronger relevance between the real time measurements and the records in signal database, the more classification votes one position that related to the record may get. In most instances, the measurements are not measured right in a reference point, but in a small region surrounded by several points. In this case, although the SVM achieves a great improvement when compared with the linear methods, defects still exist in this one-time decision strategy. Thus we use a Weighted K-Nearest Neighbor (WKNN) algorithm to do the further classification and build the combined algorithm SVM-WKNN to smooth the data and improve the performance. Experiments show that the proposed SVM-WKNN algorithm can achieve an excellent positioning accuracy and outperform the case that SVM used alone.