Evaluating polynomial regression based data aggregation in body area networks

In recent years, more and more biological sensors are being connected wirelessly. This makes them much more versatile and allows a greater degree of mobility to the wearer. However, there are downsides to the move to wireless. In wireless body area sensor network (WBASNs), power consumption and storage limitations are major obstacles that need to be addressed. Increasing capacity often means increasing weight and cost, especially when batteries are concerned. Therefore, it is important to minimize the power and storage requirements as much as possible. One approach to this is through aggregation. By having sensors report their data to a cluster head (CH), their information can be combined and volume of transmitted data reduced. Less data means more information can be stored before that data needs to be offloaded. In addition less data means fewer packets need to be sent increasing the life of the sensor. Polynomial regression can be used to create a formula that approximates this information. This paper assesses the effectiveness of this type of regression as applied to biological data of patients. This is done first by measuring the degree of compression (compression ratio). Then we evaluate the accuracy of the polynomial (correlation coefficient) for different types of patients' data. Lastly it compares these results to existing models in order to assess its effectiveness as an aggregation technique.