Extracting Gait Parameters Extracting Gait Parameters from Raw Data from Raw Data Accelerometers Accelerometers André DIAS a,b,c , Lukas GORZELNIAK b , Angela DÖRING c , Gunnar HARTVIGSEN a,d , Alexander HORSCH b,d a Norwegian Centre for Integrated Care and Telemedicine, University Hospital of North Norway, Tromsø, Norway b Institut für Medizinische Statistik und Epidemiologie, Technische Universität München, Germany c Institute of Epidemiology, Helmholtz Zentrum München, Germany d Computer Science Department, University of Tromsø, Norway
Introduction Force plates • Gait parameters are important for Gait impairment assessment http://amti.biz/ • Recovery therapy Video analysis • Several Methods – Force plates – Pressure activated sensors – Motion analysis from video http://www.mar-systems.co.uk MIE 2011, Oslo
Motivation • All existing methods are only feasible in controlled settings [setup, cost, labour] • Accelerometers have reached a stage where high frequency raw data is possible – Cheap & easy method to estimate gait parameters MIE 2011, Oslo
Goal • Can we extract basic parameters from raw data collected with accelerometers? • How do the results compare to a de- facto standard? MIE 2011, Oslo
Methods - Material • GAITrite walkway (de-facto standard) • GT3X – 30 Hz, tri-axial – Both Legs, Both Arms MIE 2011, Oslo
Methods - Subjects • KORA-Age study • 70 subjects * 4 walks each – Normal walk – Slow walk – Running – Performing mental task (counting backwards) MIE 2011, Oslo
Software • First step: extract gait parameters from GaitRite data – Closed source, unstable software provided by GaitRite – Develop raw data processing tool • Second step: estimate gait parameters from accelerometer data – Filters and peak detection MIE 2011, Oslo
Results - GT3X • Visual indications MIE 2011, Oslo
Results – Gait parameters • Algorithm Selles et al 2005 * • Second order Butterworth filter with cut off frequencies * Selles et al.; Automated Estimation of Initial and Terminal Contact Timing Using Accelerometers; Development and Validation in Transtibial Amputees and Controls; IEEE Tran on neural systems and rehabilitation eng, vol. 13, no. 1, 2005 MIE 2011, Oslo
It works... with a few subjects. But for most is a miserable failure MIE 2011, Oslo
Conclusions • Promising approach – Significant room for improvement • Sampling frequency still not enough? • More work on algorithms? • Future work – Algorithm from Jung-Ah Lee et al 2010 – E-ar sensor [located in the ear, natural balance “centre”] MIE 2011, Oslo
Questions? • andre.dias@uit.no Thanks to Matej Svedja, Jennifer Reinelt; Moritz Fuchs for the valuable help. This research was funded/supported by the Graduate School of Information Science in Health (GSISH) and the Technische Universität München Graduate School. A. Dias is supported by scholarship SFRH/BD/39867/2007 of the Portuguese Foundation for Science and Technology and Research Council of Norway Grant No. 174934. MIE 2011, Oslo
It works... with a few subjects. But for most is a miserable failure MIE 2011, Oslo
Results • Algorithm from &&&& – $$$$ filter + peak detection MIE 2011, Oslo
Conclusions • Promising approach – Significant room for improvement • Sampling frequency still not enough? • More work on algorithms? • Future work – Algorithm from ???? – E-ar sensor [located in the ear, natural balance “centre”] MIE 2011, Oslo
Recommend
More recommend
