Abstract
As wireless spectrum availability becomes increasingly important in both military and civilian applications, cognitive radios (CR) are poised to become the radio technology of choice. An important aspect to the performance of CRs is automatic modulation classification (AMC): the ability to accurately and automatically determine the modulation scheme of a received signal. Traditional methods rely on a priori and expert-based knowledge of the wireless channel and incoming signals; however, this is not generalizable to situations where the channel is severely impaired or even unknown. Deep learning (DL) has seen widespread success in fields such as image processing. In the past few years, its application to AMC has started being explored, as DL requires no a priori or expert-based knowledge. The few currently available DL applications to AMC suffer from long training times due to increasing deep layers for improved classification accuracy. This thesis proposes the use of induced class hierarchies to break down the AMC task into smaller components, while still creating deep architectures for improved classification accuracy. A publicly available synthetic radio dataset is used, which models severe channel impairments under various signal-to-noise ratio (SNR) levels. Two hierarchical convolutional neural network (CNN) models are developed, a two-level hierarchy and a three-level hierarchy. The two-level model achieves a 4% improvement in classification accuracy over the baseline model, while the three-level model achieves comparable classification accuracy. However, the training times of both models are significantly reduced, with a 2x improvement and 1.4x improvement in the two-level and three-level models, respectively. The effectiveness of induced class hierarchies in CNN classifiers is compared to a traditional AMC k-nearest neighbors (kNN) classifier, and future models are explored to mitigate the effects of error propagation.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.