Research On Context-aware Techniques Based On Channel State Information

With the rapid development of IOT technology,the context-aware techniques are gradually incorporated into people's daily lives,with the purpose of enabling computing devices to sense the surrounding environment through data collection and analysis so as to provide more intelligent services.The environment here includes both user's physical environment information,such as network conditions,interior layout,etc.,and user's own status information,such as user's location and actions.WiFi can be used for sensing as the received signal can be identified as the reflection of the environmental characteristics.Traditional WiFi-based sensing techniques are mainly implemented by RSSI(Received Signal Strength Indicator),which is determined by its convenience and easy accessibility.However,they are suffering from coarse granularity and poor time stability of RSSI.In this thesis,we deep into the physical layer to extract CSI(Channel State Information)and realize more robust and sophisticated context-aware techniques.The thesis elaborates the basic idea of CSI-based sensing technology and summarizes the advantages of CSI compared to RSSI.Aiming at solving the problems of traditional RSSI-based sensing technology,this thesis takes two typical scenarios as examples and proposes the CSI-based omnidirectional human detection and network segmentation.The former extracts three channel fingerprints,which are Rician-K,delay spread and Doppler spread.We use K-means clustering algorithm to achieve network segmentation.The latter uses PCA(Principal Component Analysis)to extract the CFR amplitudes as the feature and utilizes EMD(Earth Mover's Distance)algorithm to determine the direction of the intruder.Finally,we analyze the performance of the two applications and a large number of experiments are conducted to verify their performance.