Computational model for the recognition of lower limb movement using wearable gyroscope sensor

Human activity recognition (HAR) using inertial sensors has enabled many applications in different fields, especially healthcare and biomedical engineering. In this regard, an activity recognition system is proposed using the signals of a single gyroscope sensor placed at the shank. Principal component analysis method was utilised to exclude the redundant features from the feature set. Furthermore, different classifiers such as probabilistic neural network, k-nearest neighbour (KNN) and support vector machine (SVM) were used for recognition of walking activities. K-fold cross validation and four performance parameters namely accuracy, sensitivity, specificity, and Matthew's correlation coefficient were used to inspect the performance of the recognition model. The proposed model yielded encouraging recognition accuracy of 98.7% compared to the existing activity recognition systems. It is realised that the proposed system will potentially be utilised in the control of lower limb prosthesis and be useful tool for the gait analysis applications.