Spectrum Sensing

A non-coherent detection method that measures the energy of a received signal over a specific time-frequency block and compares it against a noise-dependent threshold. It is the most common technique due to its low computational complexity and the fact that it requires no prior knowledge of the primary user's signal characteristics.

• Mechanism: Computes the squared magnitude of FFT bins and integrates over time
• Key Weakness: Performance degrades catastrophically under noise uncertainty; cannot distinguish between a primary user signal and a burst of interference
• Threshold Sensitivity: Requires precise noise floor estimation, which is notoriously difficult in dynamic RF environments

The optimal coherent detection method when the secondary user possesses complete a priori knowledge of the primary user's waveform, including modulation type, pulse shaping, and synchronization parameters. It maximizes the received signal-to-noise ratio (SNR) by correlating the incoming signal with a known template.

• Mechanism: Convolves the received signal with a time-reversed replica of the known pilot or preamble
• Advantage: Achieves detection with minimal samples and operates reliably at very low SNR
• Limitation: Requires dedicated circuitry per waveform type; impractical for detecting unknown or encrypted signals

Exploits the inherent periodic statistical properties of man-made communication signals, which exhibit spectral correlation at specific cycle frequencies due to modulation, carrier, and framing structures. Unlike stationary noise, modulated signals generate non-zero cyclic autocorrelation.

• Mechanism: Computes the Spectral Correlation Function (SCF) to identify unique cycle frequencies
• Key Advantage: Robustly distinguishes between primary user signals and stationary noise, even below the noise floor
• Trade-off: High computational cost due to the 2D SCF calculation; can identify the specific modulation type of the detected signal

Leverages known preambles, pilot tones, or midambles embedded in standardized communication protocols (e.g., Wi-Fi preambles, LTE cell-specific reference signals) for reliable detection. This technique performs pattern matching against these deterministic sequences.

• Mechanism: Cross-correlation of the received signal with a locally generated copy of the known sequence
• Application: Highly effective in standards-compliant environments like detecting 802.11 OFDM preambles or LTE CRS
• Limitation: Protocol-specific; fails when encountering non-standard or proprietary waveforms without known patterns

A blind detection method derived from random matrix theory that analyzes the eigenvalues of the sample covariance matrix of the received signal. It exploits the fact that the presence of a correlated primary user signal alters the eigenvalue distribution compared to uncorrelated white noise.

• Mechanism: Computes the ratio of maximum to minimum eigenvalue (MME) or average to minimum eigenvalue (EME) and compares against a theoretically derived threshold
• Key Advantage: Completely immune to noise uncertainty; requires no prior knowledge of signal or noise power
• Cost: Requires multiple antenna elements or oversampling to construct the covariance matrix

Utilizes convolutional neural networks (CNNs) or recurrent neural networks (RNNs) trained on raw IQ samples or spectrograms to learn hierarchical features that distinguish primary user signals from noise and interference without explicit feature engineering.

• Mechanism: A trained model classifies each sensing window as 'occupied' or 'vacant' based on learned representations of the RF environment
• Advantage: Automatically learns robust features that generalize across varying channel conditions and unknown signal types
• Consideration: Requires substantial labeled training data and may suffer from adversarial vulnerability; computationally intensive for real-time edge deployment

An intelligent wireless system that uses spectrum sensing as its primary awareness mechanism to detect changes in the RF environment. The CR's cognition cycle begins with sensing, proceeds through analysis and reasoning, and culminates in adaptive action—adjusting transmission power, modulation schemes, or operating frequency in real-time. Without accurate spectrum sensing, a cognitive radio is blind and cannot fulfill its mandate to avoid interference.

The central objective of spectrum sensing: determining whether a licensed incumbent transmitter is active on a specific channel. Detection must be robust against the hidden node problem, where a secondary user is shadowed from the primary transmitter but can still cause interference to a nearby primary receiver. Key detection methods include:

• Matched filter detection: Maximizes SNR when PU signal characteristics are known
• Energy detection: A blind method requiring no prior knowledge of the PU signal
• Cyclostationary feature detection: Exploits periodic statistical properties of modulated signals

A spectrum hole (or white space) is a frequency band assigned to a primary user that is temporally and geographically unoccupied at a specific location and moment. Spectrum sensing must distinguish true holes from deep fades and noise uncertainty. The sensing-throughput tradeoff dictates that longer sensing durations improve detection probability but reduce the time available for secondary data transmission, requiring optimization of the sensing period.

The mathematical framework that accurately models the uncertainty inherent in spectrum sensing. Unlike a standard MDP, a POMDP acknowledges that the cognitive radio cannot directly observe the true channel state—it only receives noisy sensor readings. The agent maintains a belief state, a probability distribution over possible channel occupancies, and updates it with each new sensing observation. This framework is essential for designing optimal sensing and access policies under real-world uncertainty.

An advanced sensing technique that exploits the periodic statistical properties of modulated signals. Most man-made communication signals exhibit cyclostationarity due to carrier frequencies, symbol rates, and cyclic prefixes, while noise is typically stationary. By analyzing the spectral correlation function, this method can distinguish between different modulation types and detect signals at very low signal-to-noise ratios (SNR) where energy detection fails, though at higher computational cost.

A paradigm where multiple spatially distributed cognitive radios share their individual sensing observations to overcome the limitations of single-node detection. By combining data through hard decision fusion (OR, AND, majority rules) or soft decision fusion (likelihood ratios), the network mitigates the hidden node problem and multipath fading. The tradeoff involves increased reporting channel overhead and vulnerability to spectrum sensing data falsification (SSDF) attacks by malicious nodes.