Study on Effective CSI Feature Extraction and Deep Learning Based Indoor Positioning Method

In recent years, the indoor positioning solution market has been expanding, and there is a growing demand for indoor positioning technologies. Consequently, various techniques for the indoor positioning are explored. Traditionally, received signal strength (RSS) is studied as a simple method. However, the positioning accuracy is deteriorated significantly due to the obstruction. To solve this problem, the elements of the channel state information (CSI) are utilized. Moreover, the improved performance is obtained by using the deep learning techniques such as the deep neural networks (DNN) and the convolutional neural networks (CNN). However, the optimality is not indicated in these methods. Therefore, in this study, we propose the effective CSI feature extraction and the deep learning based indoor positioning method.