Blind standard identification with bandwidth shape and GI recognition using USRP platforms and SDR4all tools

Hongzhi Wang, Wassim Jouini, Amor Nafkha, Jacques Palicot, Leonardo S. Cardoso, Mérouane DebbahCrownCom 2010 · DOI

Introduction

In Cognitive Radio (CR) systems, devices must be "context-aware" to optimize spectrum usage and minimize interference. A critical component of this awareness is the ability to identify wireless standards "blindly"—that is, without prior coordination or signaling from the network. While traditional radios are pre-configured for specific standards, a blind receiver must infer the signal type (e.g., GSM, UMTS, or OFDM) directly from the received electromagnetic waves. This task is increasingly difficult as modern standards often share similar spectral footprints, making simple power measurements insufficient for discrimination.

Methodology: The Blind Standard Recognition Sensor (BSRS)

The researchers propose a two-stage Blind Standard Recognition Sensor (BSRS) architecture to address the limitations of simple frequency-domain detection.

The three-step Blind Standard Recognition Sensor (BSRS) architecture.

The process involves three primary steps:

  1. Band Adaptation: The incoming signal is processed to adapt to the target frequency bands.
  2. Detection Modules:
    • Bandwidth Shape Recognition: The system calculates the Power Spectrum Density (PSD) of the signal. This "spectral fingerprint" is processed by a Radial Basis Function Neural Network (RBF-NN), which compares the shape against known standards using a Mean Squared Error (MSE) metric.
  3. Guard Interval (GI) Detection: To differentiate between standards with overlapping spectral shapes (like different OFDM systems), the system employs cyclostationarity. It uses cyclic autocorrelation to detect the periodic "buffer" or Guard Interval unique to OFDM waveforms.
  4. Fusion: The outputs from various sensors (bandwidth, GI, carrier frequency, etc.) are fused to produce a final standard identification.

Experimental Setup

The system was implemented using the SDR4all software toolset, which bridges Matlab-based signal processing with Universal Software Radio Peripheral (USRP) hardware. Due to hardware constraints of the USRP v1, the researchers generated "standard-like" signals for validation:

  • GSM-like: GMSK modulation with a Gaussian filter.
  • UMTS-like: QPSK modulation with Root Raised Cosine (RRC) pulse-shaping.
  • WiFi-OFDM-like: 64 subcarriers with 16-QAM modulation.

These signals were transmitted over-the-air in the 2.4 GHz ISM band to test the system in a real-world, albeit controlled, environment.

Results and Discussion

The proposed method demonstrated high reliability in both simulations and hardware experiments. The bandwidth shape detection module achieved a false alarm probability of less than 0.5%.

Rate of successful identification as a function of SNR (dB).

As shown in the performance curves, the rate of successful identification improves significantly as the Signal-to-Noise Ratio (SNR) increases. In experimental trials using USRP platforms, the system achieved a 100% detection probability for all tested signal types once the SNR reached 0 dB or higher. The inclusion of the GI sensor proved vital as a "tie-breaker" for signals that appeared identical in the frequency domain but possessed distinct temporal structures.

Conclusion

This study demonstrates that robust, blind standard identification is achievable using affordable SDR hardware and sophisticated signal processing techniques. By combining neural-network-based spectral analysis with cyclostationary feature detection, the BSRS system provides a holistic view of the RF environment. Future work will focus on expanding the database of supported standards and testing the system's resilience against multi-path fading in more complex, non-laboratory environments.

Summary updated

BibTeX
@inproceedings{wang2010blind,
  author = {Hongzhi Wang and Wassim Jouini and Amor Nafkha and Jacques Palicot and Leonardo S. Cardoso and Mérouane Debbah},
  title = {{Blind standard identification with bandwidth shape and GI recognition using USRP platforms and SDR4all tools}},
  booktitle = {CrownCom},
  year = {2010},
  doi = {10.4108/icst.crowncom2010.9181},
}
Figure