Gait partitioning using minimal sensor data during intrinsically driven transitions

The basis for identifying stance and swing phase to classify gait requires the accurate estimation when major gait events such as initial contact and toe off occur. This study examines the ability of the Beta Process Autoregressive Hidden Markov Model (BP-AR-HMM) to estimate initial contact and toe off utilizing minimal sensor data in a non-laboratory setting. Data were sampled from a single inertial measurement unit (IMU) located on the dorsum of the foot. Participants completed average speed trials where they self-selected transitions between walking and running. Results from this study indicate that BP-AR-HMM outputs may be used to estimate gait events for a range of speeds and locomotion types.