A novel radiometric signature identification scheme for indoor environment

Abstract

With an always changing technological world, the advancement of inexpensive and basic Wireless Sensor Network (WSN) devices can possibly replace many different technologies for low energy consumption and low data rate environmental monitoring, smart homes and security applications. Radio frequency devices have the characteristic of reporting their signal level through radio signal strength indicator (RSSI). In this dissertation, changes in RSSI levels of IEEE 802.15.4 transceivers are used as indicators of the presence of an individual in indoor environment. Furthermore, with the proposed scheme, the idea of determining the identity of a particular individual by the increase of the number of devices that send their RSSI levels is introduced. To prove the concept of utilizing radiometric fingerprinting for human identification in indoor condition, raw RSSI data is processed first using filters to remove out of range data. Then statistical profiles are created from all the participating vertically aligned transmitters and associated with each human individual using the indoor space. Then a matching decision is determined after the newly generated statistical profiles are then compared with pre-stored statistical profiles (e.g., RSSI samples of same size referred to a radiometric signatures in this dissertation) in the database and a decision of matching or mismatching is given in view of RMSE threshold. To test the newly developed methodology, sets of tests were performed on total of 27 subjects for 4 predetermined static positions in Communications Protocols Laboratory in Block E-2, Kulliyyah of Engineering. The scheme was tested for two groups that emulate members of family and office groups in addition to testing based on the position of the subject in relation to the link. The outcomes demonstrated a sensitivity and specificity of 100% for a family testing group of 4 members and 5 members office group. On account of testing the best positions for subjects that gives the best performance, sitting positions showed better performance with 100% accuracy for both in-link and out-of-link positions for a resolution of 7 transmitters using the optimum decidability model. While for standing positions, the in-link position achieved 0% false positive alarms for both models and only 2.4% false negative alarms using optimum decidability model and 3.6% false negatives for the in-order of heights model. Such performance, uncovers the capability of the proposed radiometric scheme in security and home automation applications where real-time identification of occupant is a demand. Likewise, this work is an important step in empowering non-invasive recognition systems that make sensor networks more widely accepted.