Statistics-based technique for automated detection of gait events from accelerometer signals

To control an intelligent knee prosthesis for above-knee amputees, an algorithm is developed to detect gait events directly from accelerometer signals captured on the prosthesis. Using this technique, several events are automatically detected along the gait cycle. The simplicity and effectiveness of the technique is demonstrated, showing automated adaptability even for amplitude and frequency variations in gait pattern, while solving problems inherent to calibration such as offsets and scale factors. Results are equally applicable to intact limbs and further applications are also possible for events detection on periodic signals with spikes.