Establish an Innovative ML-enabled Human Activity Recognition through Edge Computing

The use of a computer system to recognize human activity or gestures, Human Activity Recognition (HAR), enables real-time behaviour and emotion prediction in individuals. In the fields of ubiquitous edge computing (EC), human behaviour analysis, and human-computer interaction, HAR is a crucial area of research. Machine learning algorithms are used in various fields of research to identify both basic and complicated human activities. In this study, we proposed a novel Garra rufa fish-optimized multi-kernelized support vector machine (GRFO-MSVM) for the recognition of human activity. The suggested approach utilizes the EC’s capabilities to train a model, which is then applied to less potent edge devices for recognition. In this study, we collected a WISDM dataset for human activity. The pre-processed data are extracted using linear discriminate analysis (LDA) to reduce dimensionality in sensor data. The proposed method is implemented using Python software. The proposed method is compared to other existing algorithms. The study indicates the SVM can improve accuracy while performing HAR within the edge. The experimental result shows the SVM achieved better performance in accuracy (98.73%), precision (98.71%), recall (98.72%), and F1-score (98.8%).