A Hands-free Smart Devices Control System Based on an Evoked Response Brain-Computer Interface and Subtle-Gesture Recognition on a Head-Worn Display
Author(s)
Peynabard, Kiavosh
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The IoT market's rapid growth, especially in homes, could reach $11.1 trillion by 2025, with 67 billion devices. However, current controlling systems for these devices are problematic for disabled users and inconvenient for others while other convenient approaches such as voice interface raise privacy issues. This paper suggests a hands-free control approach using subtle head gestures and a brain-computer interface (BCI). The proposed system uses an EEG headband to detect the selected object through flashing indicator lights on devices. This triggers a menu on a head-worn display, helping users to execute commands via head gestures. The methodology involves a low-cost EEG system, strategic stimuli placement, and data analysis using Fast Fourier Transform (FFT).
Experiments, both offline and online, were conducted to evaluate the system. Offline tests achieved an average accuracy of 96% while online studies resulted in an average of 42.5%. Integrating BCI with gestures showed promising performance, despite minor challenges like false positives. The signal strength and size issues with device small indicator lights were encountered, which reduced the quality of the signal and resulted in several false positives. Additionally, flashing lights at a low frequency might trigger seizures for epileptic users. Hence, future research could focus on advanced data analysis, using faster-flashing lights, and optimizing the display design.
In summary, this paper proposes a hybrid approach for hands-free device control using subtle head gestures and BCI. While facing initial challenges, further improvements could enhance accuracy, robustness, and user experience, advancing IoT device accessibility and usability.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis