Single sensors monitor motion

A matchbox-sized wireless sensor uses 3D accelerometers, gyroscopes and magnetometers to detect human motion down to mm accuracy for the diagnosis of illness, motion and posture capturing, gaming and animation, as well as an alternative to having to work with cumbersome and expensive video systems. A team led by Dr John Hart at the Movement Science Group at Oxford Brookes University has been using it as the core of a system called 'DataGait', which analyses how somebody walks. Parameters measured include: step time, walking rhythm, step lengths for each leg, stride length, average walking speed and Froude number, which indicates optimal energy expenditure. The sensor is the Philips 'Pi-Node' miniature attitude and heading reference system developed by the MiPlaza part of Philips Electronics. This uses 3D gyroscopes and accelerometers to determine movement in terms of acceleration and rate of turn while gravity and magnetic field sensing establish reference orientation during each sample. The information is transmitted wirelessly at 100Hz at a data rate of 250kbit/s. Inertial data is computed at up to 512Hz. A sensor fusion algorithm copes with temporary magnetic disturbances and short term accelerations. A magnetic field mapping routine corrects for hard and soft iron effects. The device weighs 35g and is free of cables, as it has an internal lithium ion battery that provides eight hours continuous use and is rechargeable from the mains. Using the input from the sensor, double integration of vertical acceleration data is used to determine the relative change in position of the centre of mass of human volunteers. In the first trial, five volunteers were asked to walk three times though a calibrated volume at walking speed, Synchronised data at 100Hz was collected both from the inertial measurement system and a Qualysis video capture system, which is the 'Gold standard' for making such measurements. Later studies showed that the method is accurate for Pakinson's disease. Currently, the research group is in the process of validating the sensors and gait models in other neurological conditions such as Huntingdon's disease, multiple sclerosis and muscular dystrophy. Out of the box comes a leaflet for self tuition, a software disc, a laptop computer, the sensor, a USB 1.1 receiver and double-sided adhesive tape to stick the sensor to the patient's back at the base of the spine. Patients are invited to walk about 10m, stop and then walk back. Total assessment time per patient is 15 minutes. Using video motion capture, a specialist laboratory assesses an average of about 1.3 patients per day. Angular resolution of the sensor is less than 0.01°, static heading accuracy is 3°, static roll and pitch accuracy is less than 1° and dynamic accuracy is 7°. Range is typically more than 10m, operating voltage is 3.6v and power consumption, 300mW. All the software to interpret the data from the inertial units was written using National Instruments LabView. Design Pointers • System uses an inertial sensor attached to a human being's sensor of mass in order to determine how they are walking and performing human tasks, as a faster and much lower cost alternative to a video capture system • Measurements made by the two systems match closely • Wheel motors for small electric vehicles exist that are based on torque motors combined with differentials • They have the potential to allow the low cost diagnosis of neurological illnesses prior to their causing irreversible damage