Ultra-Wideband Trained Artificial Neural Networks for Bluetooth Proximity Detection in Small Crowded Areas

Debnath, Satinath

Ultra-Wideband Trained Artificial Neural Networks for Bluetooth Proximity Detection in Small Crowded Areas

dc.contributor.advisor	Dr Kyle O'Keefe
dc.contributor.author	Debnath, Satinath
dc.contributor.committeemember	Hongzhou Yang
dc.contributor.committeemember	Dr Farnaz Sadeghpour
dc.date	2023-11
dc.date.accessioned	2023-09-07T19:28:09Z
dc.date.available	2023-09-07T19:28:09Z
dc.date.issued	2023-08-31
dc.description.abstract	Estimating the distance between indoor users is increasingly important in unexpected ways. One specific example is the need for electronic contact tracing demonstrated during the recent global pandemic. Smartphones are now routinely equipped with Bluetooth Low Energy radios, among other sensors, and these can be used for proximity detection based on received signal strength that is subject to errors due to poor modelling of the indoor propagation environment. Some high-end smartphones have now also been equipped with ultra-wideband ranging radios that provide a much more precise range measurement. This thesis demonstrates the concept of using a limited number of UWB-equipped smartphones to gather data to train Artificial Neural Networks (ANN) to improve short-range distance estimation among Bluetooth users. The trained RSSI to range model can be used for proximity determination by other Bluetooth users in small, crowded areas. Two ANN algorithms were trained using RSSI measurements from three BLE advertising channels and UWB range as ground truth and training data. The initial training and testing were conducted in a semi-empty office laboratory with 2130 observations. The RF model used 1917 samples (90% of data) for training and 213 samples (10%) for testing, while the CNN method used 1704 samples (80% of data) for training and 426 samples (20%) for evaluation. The trained neural network models were tested in two other office environments under different user conditions. The results indicate that the ANN models can estimate proximity in a new environment without further training with a mean error of less than 1.2 metres, within a range of up to 6 metres at line-of-sight (LOS). In highly constrained non-line-of-sight (NLOS) areas in the first office room, the proposed models provided proximity accuracy better than 2.9 metres. Furthermore, during testing across two adjacent office environments, each containing a single BLE device with complex furniture arrangements, the ANN models showed the proximity between the BLE devices with an error of less than 2-3 metres.
dc.identifier.citation	Debnath, S. (2023). Ultra-wideband trained Artificial Neural Networks for Bluetooth proximity detection in small crowded areas (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.
dc.identifier.uri	https://hdl.handle.net/1880/116962
dc.identifier.uri	https://dx.doi.org/10.11575/PRISM/41806
dc.language.iso	en
dc.publisher.faculty	Graduate Studies
dc.publisher.institution	University of Calgary
dc.rights	University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.
dc.subject	Bluetooth
dc.subject	Ultra-wideband
dc.subject	Proximity
dc.subject.classification	Engineering
dc.title	Ultra-Wideband Trained Artificial Neural Networks for Bluetooth Proximity Detection in Small Crowded Areas
dc.type	master thesis
thesis.degree.discipline	Engineering – Geomatics
thesis.degree.grantor	University of Calgary
thesis.degree.name	Master of Science (MSc)
ucalgary.thesis.accesssetbystudent	I do not require a thesis withhold – my thesis will have open access and can be viewed and downloaded publicly as soon as possible.