Wideband blind signal classification on a battery budget

Ramesh Harjani, Danijela Cabric, Dejan Markovic, Brian M. Sadler, Rakesh K. Palani, Anindya Saha, Hundo Shin, Eric Rebeiz, Sina Basir-Kazeruni, Fang Li Yuan

Research output: Contribution to journalArticlepeer-review

4 Scopus citations


A wideband signal sensor is an essential component to enable cognitive radio and dynamic spectrum access techniques, providing real-time detection and modulation classification in a wideband environment of interest. The problem is challenging, requiring a processing suite incorporating detection, estimation, and classification, with stringent power objectives to enable widespread use in untethered battery powered devices. This article provides an overview of an integrated system-on-chip extremely low-power solution, including a wideband mixed-signal front-end, an algorithm suite that incorporates a blind hierarchical modulation classifier, and an ASIC implementation that employs dynamic voltage-frequency scaling and parallel processing that achieves measured energy efficiency ranging between 11.9 GOPS/mW and 13.6 GOPS/mW for full channel feature extraction, resulting in power consumption of 20.1∼22.6 mW depending on the number of signals and signal bandwidth. The system bandwidth is selectable at 5, 50, and 500 MHz; in the 500 MHz case an efficient analog 8-point FFT channelizer relaxes the A/D requirement. The sensor can blindly detect and process up to 32 concurrent non-overlapping signals, with a variety of signal characteristics including single-vs. multi-carrier discrimination, carrier detection and estimation, and modulation classification.

Original languageEnglish (US)
Article number7295481
Pages (from-to)173-181
Number of pages9
JournalIEEE Communications Magazine
Issue number10
StatePublished - Oct 1 2015

Bibliographical note

Publisher Copyright:
© 1979-2012 IEEE.


Dive into the research topics of 'Wideband blind signal classification on a battery budget'. Together they form a unique fingerprint.

Cite this