Human recognition using wireless router signal

Abstract

Human identi cation technology can revolutionize numerous sectors in human life and a large number of methods already exist to identify humans such as voice recognition, ngerprint identi cation, face recognition and so on. As WiFi devices have become an inseparable commodity in our daily life, we are presenting a system which can identify human uniquely using WiFi signals and Channel State Information( CSI). Every person has some unique moving features and gestures which can be predicted by WiFi spectrum sensing. When a person walks through a region that is emitting WiFi transmission he or she can be easily identi ed by our model. Every person moves in a unique manner and therefore causes unique disturbances in the WiFi signals. Using Channel State Information (CSI)of the Wi-Fi signal, we have extracted 10 uncommon characteristics that separate one human being from another. We have analyzed channel state properties of a communication link from the transmitter to receiver and their combined e ects. In our database, we stored the trajectory of di erent people and matched them against measured trace. Our system has showcased 93% to 83% accuracy for K-NN, 94.09% to 88.15% for SVM and 96.05% to 89.84% for MLP for a group of 10 to 50 people. Our system has also shown an accuracy of 96% for K-NN, 97% for MLP in detecting gender for males from the 50 people and an accuracy of 86% for K-NN, 92% for MLP in detecting gender for female from 50 people consisting of 39 male and 11 female. However, the gender identi cation accuracy for both male and female were an equal 94% for KNN and 97% for MLP when the dataset consisted of 11 male and 11 female. Our proposition is that we can implement our system in residential homes and medium size o ces as smart security system for identifying humans.