The purpose of this investigation was to determine the accuracy of predicted walking and running speed and distance by smart insoles.
10 recreationally active subjects volunteered to perform two outdoor walking and running trials at speeds ranging from 1.5 - 4.8 m/s while wearing insoles containing Inertial Measurement Units - IMUs (Plantiga Technologies Platform) sampling at 416 Hz. Subjects were instructed to walk or run around a track with timing gates set 50 and 100 meters apart. Insole data was analyzed using a custom algorithm to determine instantaneous walking and running speed.
Results showed that speeds were predicted within -0.24% error (R2=1.00%, p=0.00) and distances were predicted within 1.93% error (R2=1.00%, p=0.00) (CI=95%, N=32).
Plantiga Technologies IMU’s are an accurate and simple way to predict speeds across walking and running activities. It is suggested that this technology be further studied and applied to research, high performance sports, and rehabilitation for assessment which is not confined to location or accessible equipment.
Before each trial, distance was marked with a tape measure on the straight portion of a track. Two timing gates (Smartspeed PT, Fusion Sport) were placed at the beginning and the end of a track. Cones were placed 15m before the first gate (start line) and 15m after the second gate (stop line), where subjects would start and end their trial. Continuous split times were taken as the participant moved through the two gates.
One researcher stood at the first timing gate to mark on the plantiga.io app when it was crossed (on one phone) and to run the Smartspeed app (on another phone). This was necessary to help process the data and correlate the timing gate results.
Subjects started 15m before the first gate, allowing subjects to reach a consistent travelling speed before crossing the first timing gate. Subjects were instructed to maintain consistent speed through the timing gates as best they could throughout each trial. Subjects completed one walk trial followed by one run trial. One minute rest was provided between each trial.
When the trial started, both the Smartspeed app timer (one phone) and timer on the Plantiga.io app (second phone) were activated, followed by subjects standing still for five seconds before performing each activity. When 10 minutes had elapsed, subjects stood still for five seconds and then the timer was stopped on plantiga.io. Five seconds of stillness before and after each trial allows Plantiga algorithms to accurately be deployed.
Subjects were instructed to stop at the cone, 15m past the second timing gate, when it was not feasible to complete another lap within the 10 minute period. For each subject, this procedure was repeated for a 10 minute run.
We tackle the problem by breaking it down into two sub-problems:
• Compute the velocity • Periodically correct any errors
We use a strap-down inertial navigation system utilizing a gyroscope and an accelerometer to compute acceleration, velocity and position estimates. Velocity from each foot is computed individually and then optimally combined to produce a final estimate.
Micro-electromechanical system (MEMS) IMUs are noisy and require post processing to produce accurate measurements. They also suffer from instability errors such as bias. In order to correct measurement artifacts introduced by these errors, reference data from multiple sources can be combined to produce better estimates. This is sometimes called Sensor Fusion and we implement this by means of an adaptive Kalman Filter. The filter periodically estimates errors and corrects the final measurements to produce accurate measurements.