Precision EOG: The Next Interface for IoMT-based smart home control
Source PublicationDisability and Rehabilitation: Assistive Technology
Primary AuthorsEl-Gindy, El-Shafai, Soliman et al.
The Mechanics of IoMT-based smart home control
Researchers have achieved 97.7% accuracy in translating ocular flickers into household commands, overcoming the signal-to-noise hurdles that previously limited bio-signal interfaces. This interface facilitates IoMT-based smart home control by mapping eye movements to a graphical user interface, allowing users to organise their environment without physical movement. By utilising two specific transforms, the platform filters out the bio-electrical interference that typically plagues wearable sensors. While older methods relied on basic thresholding, this approach employs the Stockwell transform and Daubechies (db4) wavelets to isolate abrupt potential changes with high fidelity.
These results were observed under controlled laboratory conditions, so real-world performance may differ.
Optimising Signal Classification
The study compared four distinct architectures: Support Vector Machines (SVM), Kernel Neural Networks (KNN), Ensemble Trees (ET), and Convolutional Neural Networks (CNN). Despite the current trend toward deep learning, the SVM model paired with the db4 wavelet provided the most reliable results. The researchers extracted statistical attributes from the processed signals to characterise each movement precisely, ensuring that involuntary blinks do not trigger accidental commands. The system successfully managed several functions:
- Environmental adjustments including temperature and lighting.
- Security protocols for doors and windows.
- Media and communication device operation.
Impact and Technical Constraints
This high-fidelity detection allows for real-time interaction with minimal latency, moving assistive technology closer to fluid human-machine integration. The use of Symlets (Sym4) wavelets also showed promise but lagged behind the db4 results. The ability to monitor and assist patients in real time reduces the burden on human caregivers. However, the study does not solve the practical challenge of electrode gel dehydration and skin impedance changes over 24-hour cycles. Future research must bridge the gap between lab-based accuracy and long-term wearable durability.