2016/05/17 by Benjamin Migliori, Riley Zeller-Townson, Migliori, Benjamin +5
Computer Science · Biochemistry, Genetics and Molecular Biology · #Wireless Signal Modulation Classification #Spider Taxonomy and Behavior Studies #RNA and protein synthesis mechanisms
paper · pdf · doi:10.48550/arxiv.1605.05239
Automatic modulation classification (AMC) is an important task for modern\ncommunication systems; however, it is a challenging problem when signal\nfeatures and precise models for generating each modulation may be unknown. We\npresent a new biologically-inspired AMC method without the need for models or\nmanually specified features --- thus removing the requirement for expert prior\nknowledge. We accomplish this task using regularized stacked sparse denoising\nautoencoders (SSDAs). Our method selects efficient classification features\ndirectly from raw in-phase/quadrature (I/Q) radio signals in an unsupervised\nmanner. These features are then used to construct higher-complexity abstract\nfeatures which can be used for automatic modulation classification. We\ndemonstrate this process using a dataset generated with a software defined\nradio, consisting of random input bits encoded in 100-sample segments of\nvarious common digital radio modulations. Our results show correct\nclassification rates of > 99% at 7.5 dB signal-to-noise ratio (SNR) and > 92%\nat 0 dB SNR in a 6-way classification test. Our experiments demonstrate a\ndramatically new and broadly applicable mechanism for performing AMC and\nrelated tasks without the need for expert-defined or modulation-specific signal\ninformation.\n