Data Separation Considering Smoothness of Sensing Data in Physical Wireless Parameter Conversion Sensor Networks

Abstract

Physical wireless conversion networks have an advantage that the fusion center can simultaneously receive each data sent from multiple sensors. However, the separation of the mixed data was difficult without sensor labels for distinguishing the data. In this paper, we aim to separate the mixed data into passes composed of continuous data points having smooth gradient by using a $K$-shortest pass algorithm. In our previous method, we used the distance between data points as the cost and calculate passes so as to obtain the minimum length, but the method prone to select a wrong pass at intersection points of the passes. In the proposal method, assuming the position of a data point changes smoothly, we use the smoothness of a pass (gradient) as the additional cost. However, it is actually difficult to calculate the gradient of a pass that has not yet been separated. Therefore, in this paper, on the basis of the assumption that the gradient can be somehow obtained, we performed data separation experiments for some mixed data. As a result, a good performance is obtained, so we report the results in this paper.