Publication

Introduction - New trends in analysis of human movement promote small low-cost portable devices and combine them with the traditional fixed lab-equipment such as force plates and stereophotogrammetry [1]. These new portable devices open for new applications, such as analysis of movement in daily situations and for new biofeedback possibilities as training assistant for rehabilitation or sportive activities [2]. However, currently available systems are often closed and/or designed for a specific application. We developed an open platform able to adapt to a number of different applications by enabling the analyst to select its sensors and customize the processing algorithms. Methods After describing a number of use cases, functional and technical requirements were derived and a first prototype device implemented. Such device is able to collect and process data in real-time from any combination of six sensors from the following list: inertial movement units (IMUs), GPS, cameras and pressure sensors. An IMU consists of a 3D accelerometer, a 3D gyroscope and a 3D magnetometer. Further, the device provides basic acoustic, visual and tactile biofeedback. Based on a Linux operating system, the software is a combination of open drivers and C code. Data is stored in a standard text format. In addition, the device is wireless capable. The device was verified outdoors with a traffic safety application designed for testing purposes. The application used the following sensors: two IMUs, GPS, one camera, one 3D accelerometer and one pressure sensor. Three participants walked a 500-m loop with four zebra crossings. A total of eight test runs were made. At each crossing, the participants were instructed whether to scan for coming traffic by turning the head left and right, or just cross the road without scanning (after the experimenter had made sure no traffic was coming). The application comprised of three algorithms to detect: 1) the presence of a crossing, 2) the initiation of gait, 3) the head scanning movement - to determine whether or not to give a warning feedback. System performances were assessed by calculating sensitivity and specificity for the three algorithms.Results - Feedback was provided with 87% success rate. Specificity and sensitivity are reported in Table 1 for the three algorithms. Gait initiation was detected in all tests in 0.4±0.2 s. Head scan was detected with 90% success rate. Crossing detection had poor performances which originated from the low GPS accuracy (8.6±4.8m). Conclusions - The portable device developed in this study collects and processes in real-time information from a custom number of sensors such as IMUs, GPS, cameras and pressure sensors. Further, this device provides basic visual, audio and tactile feedback and relies on totally open software. The openness and flexibility of such device make it suitable for many applications spanning from biofeedback application in laboratory to activity assistance in a naturalistic set-up.