Unobtrusive sensing solutions for home-based monitoring

Student thesis: Doctoral Thesis


This Thesis proposes the usage of Unobtrusive Sensing Solutions (USSs) to monitor activities and rehabilitation exercises in a simulated home environment. The main Contribution to Knowledge of this Thesis is in the design and implementation of a novel Sensor Data Fusion Architecture capable of analysing data from both homogeneous and heterogeneous SSs. Research findings from the technical assessment on Post Stroke Rehabilitation Exercises (PSRE) monitoring indicated that individual muscles unit action potentials activated during neural stimulation could be detected in targeted muscle groups using a thermal and a UWB Radar USS. Detailed studies involving USSs such as HB100 and FMCW Radar and an Infrared Thermopile Array (ITA) thermal sensors indicated the ability to obtain postures, range of motion and the velocity values of the upper and the lower extremities during PSREs. Furthermore, experimental results from a study in indoor activity monitoring indicated instances of activity recognition during tea/coffee making and the classification of the same activities using Data Mining (DM) tools with an average predictive accuracy of 95%. A case study on Sprained Ankle Rehabilitation Exercises (SPAREs) also reported an average clustering accuracy of 96.9% in the heterogeneous datasets involving thermal and Radar SS on instances of SPAREs such as dorsiflexion, plantarflexion, inversion and eversion. On fall detection, a percentage accuracy of 95% was obtained on the Cluster-Based Analysis (CBA) involving data gleaned from an ITA sensor in a simulated environment. Probabilistic analysis of the temporal and sequential aspects of human behaviour during ADLs indicated the ability to recognise abnormal behaviour during the performance of ADLs using data from contact switch sensors. The CBA of the ADLs datasets using KMA and HCA models indicated an average accuracy of more than 99% in all the data mining metrics such as Stochastic Gradient Descent, Neural Networks and Support Vector Machine, amongst others.
Date of AwardDec 2021
Original languageEnglish
SupervisorChristopher Nugent (Supervisor), Jim McLaughlin (Supervisor) & Sally McClean (Supervisor)


  • Thermal sensor
  • Radar sensor
  • Home environment
  • Data mining
  • Data analysis

Cite this