Spatio-temporal gait parameters as estimated from wearable sensors placed at different waist levels

Wearable devices are able to capture movement-related characteristics from inertial sensors integrated in them. Many spatio-temporal parameters of gait can be estimated by the acquisition of inertial data, but their accuracy depends on the placement of the devices on the body, as well as on the numerical values chosen for the estimation techniques. In this work, three inertial sensors placed at three different heights of the trunk are used to collect data from healthy adult participants walking at three different speeds. Step length and step velocity were calculated from accelerometer data. For the estimation of these parameters high-pass filtering is required: fifteen different values of the filter cut-off frequency were analyzed for the subsequent step length estimation. The results were compared against those measured from a marker-based movement analysis system. Estimation accuracy of both step length and step velocity resulted significantly affected by both sensor location and cut-off frequency of the filter. These preliminary results suggest that placing the sensor too low leads to an increased estimation error, while frequencies up to around 1 Hz lead to acceptable results, with a significant decrease in estimation accuracy above that value.