Lower extremity angle measurement with accelerometers--error and sensitivity analysis.

Closed-loop control techniques for the restoration of locomotion of paraplegic subjects are expected to improve the quality of functional neuromuscular stimulation (FNS). We investigated the use of accelerometers for the assessment of feedback parameters. Previously, the possibility of angle assessment of the lower extremities using accelerometers, but without integration, was demonstrated. The current paper evaluates and assesses this method by an error and sensitivity analysis using healthy subject data. Of three potential error sources, the reference system, the accelerometers, and the model assumptions, the last was found to be the most important. Model calculations based on data obtained by the Elite video motion analysis system showed the rigid-body assumption error to be dominant for high frequencies (greater than 10 Hz), with vibrations in the order of 1 mm resulting in errors of one radial or more. For low frequencies (less than 5 Hz), the imperfect fixation of the accelerometers combined with a nonhinge type knee joint gave an error contribution of +/- 0.03 rad. The walking pattern was assumed to be two-dimensional which was shown to result in an error of +/- 0.04 rad. Accelerations due to rotations of the segments could be neglected. The total error computed for low frequencies (+/- 0.07 rad) was comparable to the experimental difference between the current and the reference system.