Efficient WiFi Fingerprint Training Using Semi-supervised Learning

Fingerfrinting based WiFi positioning approach needs an off-line training phase to build a radio map with received signal strength indication vector of each reference point. In existing systems, this training phase may cost a tremendous amount of workload to achieve satisfying location result. To cut down on the workload notably and guarantee the location result in the meantime, we will introduce an efficient WiFi fingerprint training method: Fa-Fi namely fast fingerprint generation, which uses semi-supervised learning in this article. This proposed method can reduce the training phase time cost to about 1/5, and guarantee the localization accuracy at the same time.