Evaluating C-SVM, CRF and LDA classification for daily activity recognition

Abstract

The ability to recognize human activities from sensed information becomes more attractive to computer science researchers due to a demand on a high quality and low cost of health care services at anytime and anywhere. This work compares C-Support Vector Machine (C-SVM), Conditional Random Fields (CRF) and Linear Discriminant Analysis (LDA) for imbalanced dataset to perform automatic recognition of activities in a smart home. This comparative study offers a guideline for choosing the appropriate algorithms for automatic recognition of activities. We conduct several experiments carried out on real world dataset and show that the results obtained with C-SVM are very promising. C-SVM is able to correct the inherent bias to majority class and yields improvement in the class accuracy of activity classification (75.5%) in comparison with CRF (70.8%) and LDA (72.4%) methods.