Smartphone Sensors Based Physical Life-Routine for Health Education

Abstract

The physical and the mental health of a human being largely depends upon his physical life-routine (PLR) and today’s much advanced technological methods make it possible to recognize and keep track of an individual’s PLR. With the successful and accurate recognition of PLR, a sublime service of health education can be made copious. In this regard, smartphones can play a vital role as they are ubiquitous and have utilitarian sensors embedded in them. In this paper, we propose a framework that extracts the features from the smartphone sensors data and then uses the sequential feature selection to select the most useful ones. The system employs a novel approach of codebook assignment that uses vector quantization to efficiently manipulate the data coming from the smartphone sensors of different nature and serve as a data compression module at the same time. The proposed system uses a multilayer perceptron classifier to differentiate among different PLRs. The experimentation was performed on the benchmark Real-life HAR dataset. It provides the data of four sensors: accelerometer, gyroscope, magnetometer, and global positioning system (GPS) for the recognition of four activities namely active, inactive, walking, and driving. The performance of the proposed system was validated using 10-fold cross-validation and the confidence of the system was recorded to be 91.80%.