Exoskeleton-Based Multimodal Action and Movement Recognition: Identifying and Developing the Optimal Boosted Learning Approach

This paper makes two scientific contributions to the field of exoskeleton-based action and movement recognition. First, it presents a novel machine learning pattern recognition-based framework that can detect wide range actions movements - walking, walking upstairs, downstairs, sitting, standing, lying, stand sit, sit stand, lie, lie with an overall accuracy 82.63%. Second, comprehensive comparative study different approaches Random Forest, Artificial Neural Network, Decision Tree, Multiway Support Vector Machine, k-NN, Gradient Boosted Trees, Stump, Auto MLP, Linear Regression, Naive Bayes, Bayes (Kernel), Discriminant Analysis, Quadratic Deep Learning applied this framework. The performance each these was boosted by using AdaBoost algorithm, Cross Validation approach used for training testing. results show in form, k- NN classifier outperforms all other is, therefore, optimal method purpose. presented discussed uphold importance work contribute towards augmenting abilities assisted independent living elderly future Internet Things-based environments, such as Smart Homes. As specific use case, we also discuss how findings our are relevant capabilities Hybrid Assistive Limb exoskeleton, highly functional lower limb exoskeleton.