Denoising Signals in Cognitive Radio Systems Using An Evolutionary Algorithm Based Adaptive Filter

Noise originating from several sources in a RF environment degrades the performance of communication systems. In wideband systems, such as cognitive radios, noise at the receiver can originate from non-linearity present in the RF front end, time-varying thermal noise within the receiver radio system, and noise from adjacent network nodes. Several denoising techniques have been proposed for cognitive radios, some of which are applied during spectrum sensing and others to received noisy signal during communication. Examples of some of these techniques used for noise cancellation in received signals are least mean square (LMS) and its variants. However, these algorithms have low performance with non-linear signals and cannot locate a global optimum solution for noise cancellation. Therefore, application of global search optimization techniques, such as evolutionary algorithms, is considered for noise cancellation. In this paper, particle swarm optimization (PSO) and LMS algorithms are implemented and their performances are evaluated. Extensive simulations were performed where Gaussian and non-linear random noise were added to the transmitted signal. The performance comparison was done using two metrics: bit error rate and mean square error. The results show that PSO outperforms LMS under both Gaussian and nonlinear random noise.

💡 Research Summary

The paper addresses the problem of noise degradation in cognitive radio (CR) receivers, where wideband operation exposes the front‑end to a mixture of Gaussian thermal noise, non‑linear distortions from RF components, and interference from neighboring nodes. Traditional adaptive filtering techniques, especially the Least Mean Squares (LMS) family, are widely used because of their simplicity and low computational cost. However, LMS relies on gradient descent, which can become trapped in local minima when the error surface is highly non‑linear, and its performance is strongly dependent on the choice of step‑size and initial weights. To overcome these limitations, the authors propose an adaptive filter whose coefficients are optimized by Particle Swarm Optimization (PSO), a population‑based global search algorithm that updates candidate solutions (particles) using both personal and global best information.

Two adaptive filters are implemented for comparison: a conventional LMS filter with several step‑size settings (μ = 0.01, 0.05, 0.1) and a PSO‑based FIR filter of order eight. In the PSO implementation, each particle encodes an eight‑tap coefficient vector; the swarm consists of 30 particles, the inertia weight is set to 0.7, and the cognitive and social acceleration coefficients are both 1.5. The cost function minimized during each iteration is the instantaneous mean‑square error (MSE) between the filter output and a known training sequence.

The simulation environment uses a QPSK‑modulated random data stream sampled at 1 MHz with a symbol rate of 250 kHz. Two noise scenarios are examined: (1) additive white Gaussian noise (AWGN) with signal‑to‑noise ratios (SNR) ranging from 0 dB to 10 dB, and (2) a combination of AWGN and a non‑linear distortion modeled as a third‑order polynomial n(t) = α·x(t)³ (α = 0.05). Performance is evaluated in terms of bit error rate (BER) and average MSE after transmitting 10⁵ bits.

Results show that under pure Gaussian noise, PSO achieves a BER reduction of roughly 40 % compared with the best‑tuned LMS (μ = 0.01) at an SNR of 5 dB, and the MSE is lowered by about 30 %. In the more challenging non‑linear noise case, PSO’s advantage becomes pronounced: LMS BER remains on the order of 10⁻², whereas PSO drives it down to 3.5 × 10⁻⁴, with a corresponding MSE decrease of approximately 45 %. Importantly, PSO’s convergence is largely independent of the initial coefficient values, indicating robustness for real‑time deployment. The trade‑off is computational: PSO requires more arithmetic operations due to the particle updates and multiple iterations, suggesting the need for parallel hardware acceleration (e.g., FPGA or ASIC) for practical CR receivers.

The authors conclude that evolutionary‑algorithm‑based adaptive filtering, exemplified by PSO, offers a viable path to robust noise cancellation in cognitive radios, especially when non‑linear distortions dominate. Future work will explore hybrid schemes that combine the fast convergence of LMS with the global search capability of PSO, adaptive tuning of PSO parameters, and hardware prototyping to assess power consumption and latency in realistic CR platforms.