Abstract
In recent years, a plethora of algorithms have been devised for efficient human activity recognition. Most of these algorithms consider basic human activities and neglect postural transitions because of their subsidiary occurrence and short duration. However, postural transitions assume a significant part in the enforcement of an activity recognition framework and cannot be neglected. This work proposes a hybrid multi-model activity recognition approach that employs basic and transition activities by utilizing multiple deep learning models simultaneously. For final classification, a dynamic decision fusion module is introduced. The experiments are performed on the publicly available datasets. The proposed approach achieved a classification accuracy of 96.11% and 98.38% for the transition and basic activities, respectively. The outcomes show that the proposed method is superior to the state-of-the-art methods in terms of accuracy and precision.
Highlights
Human Activity Recognition (HAR) deals with the recognition, interpretation, and assessment of human daily-life activities
Wan et al [40] demonstrated a Convolutional Neural Networks (CNN) framework that showed that the fine-tuned conventional CNN still outperforms Support Vector Machine (SVM), Multilayer Perceptrons (MLP), Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) networks
The variables involved are as follows: ‘actid’ is a vector composed of activity IDs in the form of integers ranging from 1 to 5, ‘actnames’ is a vector composed of the activity names corresponding to their respective activity IDs, ‘feat’ is a feature vector composed of 60 features against every observation, and ‘featlabels’ is a list of names corresponding to every feature
Summary
Human Activity Recognition (HAR) deals with the recognition, interpretation, and assessment of human daily-life activities. Activity recognition using standard M.L approaches such as K Nearest Neighborhood (KNN) [8,9,10], Support Vector Machine (SVM) [11,12,13,14], Decision Tree (DT) [15,16], Random Forest (RF) [17,18] and Discrete Cosine Transform (DCT) [19,20,21], etc., have been reported to produce good results under controlled environments [22] The accuracy of these models heavily depends on the process of feature selection/extraction. It is worth the cost because a HAR system requires accurate classification results of the deep learning models In this context, this work proposes a hybrid deep-learning based approach where models are trained simultaneously, instead of in a pipelined setup, to recognize basic and transitional human activities. Supplementary Materials contains the repository link for the source code and datasets used in this approach
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
