Discrete Kalman Filter based Sensor Fusion for Robust Accessibility Interfaces

20th Argentinean Bioengineering Society Congress

物理学;生物科学

Ghersi, I.^1,2 ; Marino, M.^2 ; Miralles, M.T.^1,2

Pontificia Universidad Católica Argentina, LaBIS-Facultad de Ciencias Fisicomatemáticas e Ingenieria-UCA, CABA, Av. A. Moreau de Justo 1500, Argentina^1

Universidad de Buenos Aires, CIDI-Facultad de Arquitectura Diseo y Urbanismo-UBA, Intendente Guiraldes 2160, Pabellón III, Cuidad Universitaria, CABA, Argentina^2

关键词: Alternative input devices;    Complementary methods;    Discrete Kalman filters;    Discrete-time Kalman filters;    Embedded signal processing;    Human Machine Interface;    Magnetic field sensors;    Magnetometer-accelerometer;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/705/1/012036/pdf
DOI  :  10.1088/1742-6596/705/1/012036

学科分类：生物科学（综合）

来源: IOP

【 摘 要 】

Human-machine interfaces have evolved, benefiting from the growing access to devices with superior, embedded signal-processing capabilities, as well as through new sensors that allow the estimation of movements and gestures, resulting in increasingly intuitive interfaces. In this context, sensor fusion for the estimation of the spatial orientation of body segments allows to achieve more robust solutions, overcoming specific disadvantages derived from the use of isolated sensors, such as the sensitivity of magnetic-field sensors to external influences, when used in uncontrolled environments. In this work, a method for the combination of image-processing data and angular-velocity registers from a 3D MEMS gyroscope, through a Discrete-time Kalman Filter, is proposed and deployed as an alternate user interface for mobile devices, in which an on-screen pointer is controlled with head movements. Results concerning general performance of the method are presented, as well as a comparative analysis, under a dedicated test application, with results from a previous version of this system, in which the relative-orientation information was acquired directly from MEMS sensors (3D magnetometer-accelerometer). These results show an improved response for this new version of the pointer, both in terms of precision and response time, while keeping many of the benefits that were highlighted for its predecessor, giving place to a complementary method for signal acquisition that can be used as an alternative-input device, as well as for accessibility solutions.