Human Gait Modelling Using Hidden Markov Model For Abnormality Detection

This paper presents a novel approach to human gait analysis using wearable Inertial Measurement Unit(IMU) sensor-based technique. The proposed system emphasizes on detection of certain abnormal gait patterns. It includes hemiplegic and equinus gait which are synthetically generated in our lab. The designed prototype contains an IMU sensor with 3 axial accelerometer and gyroscope. It provides linear acceleration and angular velocity of human foot. A probabilistic framework, Hidden Markov Model(HMM) is applied to model bipedal human gait. This model uses Symbolic Aggregate Approximation(SAX) method for generating observation sequences obtained from sample gait cycles.The detection of abnormal gait pattern is based on maximum log-likelihood of an unknown observed sequence, generated from a gait cycle. The experimental results demonstrate that the proposed HMM-based technique is able to detect gait abnormality in gait data. The proposed personalized gait modelling approach is cost effective and reliable to implement in gait rehabilitation process.