A Surrogate Data Based Nonlinear Feature Validation Method and Its Application in Cloud Computing

Abstract

Cloud computing provides powerful processing capability and integrates data from various wearable sensors to provide online services that can increase the quality of human life and promote health awareness. Nonlinear features are efficient in the analysis and mining of the sensor data in the cloud as the bio-signals are inherently nonlinear. However, most nonlinear analysis algorithms were originally not design to cope with the noisy and non-stationary data streams from the wearable sensors. Therefore, it is a crucial step to verify the validity of the nonlinear features of the signals in the cloud before we use them. In the study, we proposed a surrogate data based nonlinear feature validation method to evaluate the validity of the nonlinear features. The experiments on two groups of signals were conducted to show the effectiveness of the method. One is computer-generated signals and the other is the skin conductance collected from a subject when she was under eight kinds of emotional states. In addition, the results of emotion classification experiment show that the surrogate data based nonlinear feature validation method is a practical way to select the valid nonlinear features of the sensor data in the cloud, and thus improve the accuracy and reliability of the classification model.