Multi-modal Deep Neural Features for Classification of Gait Abnormality

Abstract

Clinical gait analysis plays a vital role in diagnosis and monitoring neurological and musculoskeletal injuries. Qualitative gait assessment depends on subjective observations, manual measurements, and specialized equipment. Recently machine learning and deep learning based models have demonstrated significant accuracy in gait analysis. But dynamic feature extraction is always a challenging problem in temporal gait data analysis. After extracting dynamic features, a Fully-connected Neural Network (FNN) is employed to classify of gait abnormalities using GaitRec standard dataset. The proposed multi-modal features based classification model achieves 96.22% accuracy and it outperforms state-of-the-art methods