A novel hybrid framework for optimal feature selection and classification of Human Activity Recognition

Abstract

In today's world individuals health concern has improved a lot with the help of advancement in the technology. To monitor an age old person or a person with disability, now-a-days modern wearable smartphone devices are available in the market which are equipped with good collection of built in sensors that can be used for Human Activity Recognition (HAR). These type of devices generate lot of data with many number of features. When this data is used for classification, the classifier may be over trained or will definitely give high error rate. Hence, in this paper, we propose a two hybrid frameworks which gives us optimal number of features that can be used with different classifiers to recognize the Human Activity accurately. It is observed from our experiments that SVM was able to classify the HAR accurately.