A non-ZUPT gait reconstruction method for ankle sensors.

Monitoring lower body motion, especially gait pattern, using low cost Inertial Measurement Units on a daily basis is becoming critically important for the diagnosis and rehabilitation of neurological diseases. The current state of the art algorithm is to double integrate motion acceleration and compensate cumulative errors by resetting velocity signals to zero at the stance-phase of each stride. However, this method is only applicable for foot-mounted sensors. For the medically more preferable ankle-mounted position, the assumption of this zero-velocity-update (ZUPT) method does not hold. In this paper, a new non-ZUPT method is proposed. We estimated the true velocity during stance-phase, and reset velocity to the estimated value instead of zero. 10 subjects were recruited for 40-meter-level flat floor walking. The stride length estimation error was reduced to 3.58% from 13.22% on average comparing to the conventional ZUPT method on an ankle-mounted sensor. Validity of this method is further supported by stairs walking of 4 more subjects.