Sensing, and the associated analytics, is a key module of any active assistive device in physical rehabilitation. Two attributes of interest for active assistive devices in enabling human locomotion are the gait cycle parameter detection and gait Energy Expenditure (EE) estimation. Traditionally these are treated independently from a sensing/analytics perspective. Gait cycle detection is usually done with cameras/inertial sensors and gait EE is performed using metabolic measuring devices (specifically the measurement of Volume of Oxygen (VO₂)). This paper presents a novel sensing mechanism using a single Inertial Measurement Unit (IMU) sensor system placed at the ankle, to measure and estimate both gait cycle parameters and gait energy expenditure.  This proposed work supports active assistance techniques such as Functional Electrical Stimulation (FES) and Exoskeleton from a sensing/analytics perspective.

Data was collected from two volunteers (aged 22, male and healthy). The collected data is passed through a Long Short Term Memory (LSTM) algorithm to train deep neural network, to predict the gait events such as Heel Strike (HS), Heel Off (HO) and Toe Off (TO). The predicted heel strike value is 10 ms delay from an actual value. From the predicted HS value, VO₂ consumption of the corresponding person is calculated. Two subjects are taken for clinical trials and obtained error less than 20%.