Adaptive Signal Analysis System for Non-periodic Signals

This proposal aims to develop advanced signal analysis tools for utilization on non-periodic radio frequency signal sources that have the capability to detect, process, generate and classify non-periodic RF signals that do not exhibit sinusoidal characteristics. The adaptive signal analysis methods including empirical mode decomposition (EMD) and variational mode decomposition (VMD) are utilized for the construction of the adaptive features by the adaptive feature construction framework for the non-periodic signals. A deep learning based classification module will be built for the signal classification and characterizing tasks. The signal processing and generating tasks can also be performed by operating on the adaptive components derived from the above adaptive signal analysis methods. A preliminary evaluation is performed for the effectiveness of the adaptive components. The results of classification accuracy of 0.93 demonstrates the effectiveness of the adaptive components. With the further construction of the adaptive features, better performance and more complicated situations can be handled for the non-periodic signals. The ultimate goal of the proposed integrated adaptive signal analysis system is to handle ubiquitous kinds of non-periodic signals.