Finger Print based Exam Hall Authentication using Zigbee

Abstract

WiFi Fingerprint Positioning is a conventional method to indoor region determination. Existing strategies are at risk of fluctuations in WiFi signal power throughout the offline phase, resulting in inconsistently received alerts. Moreover, throughout on line positioning, there is a lack of integration with historic trajectory facts. This leads to errors in both offline fingerprint acquisition and on-line place positioning.
To scope with those problems, we advise a singular method combining normality detection in the offline segment and Location Weighted K-nearest Neighbor (LWKNN) positioning in the on line segment. In the offline segment, preliminary Received Signal Strength Indication (RSSI) samples undergo preprocessing based totally on skewness and kurtosis for normality detection. If the samples normal distribution model, their probability density function is expected using the normal distribution characteristic. If no longer, estimation takes place using the kernel density characteristic version. Subsequently, values are averaged after Kalman filtering to set up a high-precision fingerprint database. During the online positioning segment, the LWKNN algorithm is employed. Initially, the Weighted K-nearest Neighbor approach estimates the position, which is then used as functions to construct a Long Short Term Memory (LSTM) community model. The most suitable direction is decided via the least rectangular method. Finally, the obtained outputs are included with historic information from the fingerprint positioning trajectory to beautify goal positioning accuracy. Experimental consequences reveal that our indoor localization method significantly improves WiFi fingerprint localization accuracy as compared to conventional localization strategies.