[home] [Personal Program] [Help]

---

Session: Poster session I
Session starts: Thursday 24 January, 15:00

---

Bart Klaassen (University of Twente)
Leendert Schaake (Roessingh Research and Development)
Jaap Buurke (Roessingh Research and Development)
Martijn van Eenennaam (University of Twente)
Bart Koopman (University of Twente)
Hans Rietman (Roessingh Research and Development )

Abstract:
Video images, combined with additional EMG and ground reaction force data, are currently widely used in daily clinical practice for clinical gait assessment. Gait event detection, like identification of initial contacts, is usually done manually from these video recordings. Video capturing with a static camera on a tripod along the side of the walking track causes the angle between the camera and the patient to change over time by panning and tilting the video camera. This complicates automatic gait event detection and gait feature extraction. A possible solution to this problem is a following camera that is usually operated manually. However, the manual handling of the camera may introduce irregularities in video recordings. To overcome these problems, an automatic camera following system has been developed that is able to follow patients without changing the angle between camera and patient. A dolly car has been motorized and mounted on a truss rail which is parallel to the walking path of the patient (7 meters). Two cameras are mounted on the dolly car. One camera is used for the actual patient video recordings and the second camera is used for tracking the subject. A fast and reliable method for tracking the subject is by use of a selective color marker. This marker is placed on the subject. A dedicated algorithm was developed, using Matlab, to control the speed of the dolly car in a smooth and safe way. As a first validation of the system, controlled gait pattern scenarios have been carried out with three subjects and compared with Vicon motion capture system measurement data for quantifying the tracking performance. These scenarios included different walking speeds as well as irregular walking patterns commonly seen in patients at the Roessingh Research and Development centre. Tracking validation results show that the system is able to track subjects performing slow (< 1 m*s^-1), normal (about 1.4 m*s^-1) walking speeds and irregular motion patterns with a maximum offset (difference between the centre of patient and video camera/dolly car) of 585±59 mm and a minimum offset of 20±85 mm. The maximum offset produces an angular deviation of 13±1.4 degrees between the two centers (patient at 2.5 meter distance from the camera) and still makes video images eligible for clinical observations and automatic feature extractions. Higher walking speeds (> 2 m*s^-1) have different requirements for the tracking algorithm and hardware, and need further development.