This letter investigates channel capacity bounds and constellation shaping for memoryless mixed Gaussian-impulsive noise. We derive capacity bounds utilizing the entropy power inequality and the dual capacity expression. We demonstrate that these bounds become asymptotically tight, yielding a closed-form asymptotic expression for the channel capacity. Guided by this theoretical analysis, we propose a low-complexity, non-iterative constellation shaping method. Simulation results validate the tightness of the derived bounds and confirm that the proposed shaped constellation achieves the highest mutual information among the considered baselines.

Communication systems operating in mixed white Gaussian and impulsive noise (IN) environments suffer from severe performance degradation, rendering conventional constellations suboptimal. In this paper, we propose a joint geometric and probabilistic shaping scheme designed to maximize the cutoff rate (CR) under such mixed noise conditions. Instead of relying on loose approximations, we derive rigorous closed-form lower and upper bounds for the CR. To overcome the analytical intractability of the integral terms, we develop a novel piecewise linear approximation that exploits the algebraic structure of the noise model, yielding tractable objectives for optimization. Subsequently, a projected gradient-based algorithm is employed to jointly optimize the constellation geometry and probability mass function. Numerical results demonstrate that our derived bounds are tight, and the proposed hybrid shaping scheme significantly outperforms conventional baseline schemes, especially when the input power is not large enough.

Bursty mixed noise is typically characterized by the combination of white Gaussian noise (WGN) and non-Gaussian impulsive noise (IN) with memory. The performance of communication systems can experience significant degradation when conventional signal detection algorithms designed for Gaussian noise are applied in the presence of bursty mixed noise. To handle this challenge, we investigate demodulation techniques and conduct a theoretical performance analysis under bursty mixed noise. Specifically, we utilize the statistical model of channel noise to derive demodulation methods for both linear and nonlinear modulation schemes. We further analyze the theoretical error probability for M-ary PSK/QAM and provide closed-form expressions. For MSK/CPM modulations, we derive asymptotic lower and upper bounds for the error probability, accompanied by analytical expressions. Simulation results demonstrate that the proposed demodulation methods outperform baseline techniques by more than 1.9 dB when the error probability approaches 10-3. Additionally, the theoretical analysis closely aligns with the simulation results.

In many communication scenarios, the communication signals are contaminated by both Non-Gaussian impulsive noise (IN) with memory and white Gaussian noise (WGN), i.e., bursty mixed Gaussian-impulsive noise. To achieve desirable communication performance, it is necessary to establish an accurate and tractable model of the bursty mixed noise. In this paper, we propose a closed-form and heavy-tail multivariate probability density function (PDF) to model the bursty mixed noise. The proposed model is the weighted combination of the Gaussian distribution and student distribution that separately describe the background WGN and the bursty IN. We design the model parameters estimation approach based on the empirical characteristic function of the proposed noise model. The performances of parameter estimation are also analyzed. Numerical results show that our proposed bursty mixed noise model can well match the measured noise in practical scenarios. Meanwhile, the estimation performances in terms of normalized mean square error (NMSE) verify the effectiveness of presented parameters estimation methods.

Bursty mixed noise composed of impulsive noise with memory and white gaussian noise exists in many practical applications. However, existing signal demodulation algorithms are designed without considering the characteristics of the bursty mixed noise, resulting in performance degradation. Therefore, an accurate model for the bursty mixed noise needs to be built to design the corresponding signal demodulation algorithms. In this paper, we propose a novel statistical model based on the mul-tivariate student distribution for the bursty mixed noise, based on which we design the maximum likelihood (ML) demodulation method and analyze the theoretical bit error rate performance. Simulation results show that the proposed model can describe the bursty noise well and there is remarkable demodulation gain of the ML method as compared with conventional ones.

Minimum shift keying (MSK) is a continuous phase modulation with both high spectrum efficiency and power efficiency, and has been widely used in wireless communication systems. Existing demodulation algorithms of MSK signals mainly consider either white Gaussian noise (WGN) or non-Gaussian impulsive noise (IN). However, the MSK signals suffer from the mixture of WGN and IN in many practical scenarios. For these environments, directly applying these demodulation algorithms could not achieve desirable bit error rate (BER) performance. In this paper, we address the problem of demodulation of MSK signals under the mixture of WGN and IN. Based on the approximate probability density function, the Viterbi algorithm based demodulation scheme is proposed by calculating the maximum-likelihood (ML) branch metric and the corresponding theoretical BER performance is analyzed. Furthermore, several popular branch metrics are compared in terms of both BER and complexity. Simulations results show that the proposed ML branch metric can obtain the best BER performance under all the scenarios, and the theoretical BER matches the simulation results well.

Communication systems suffer from the mixed noise which contains both non-Gaussian impulsive noise (IN) and white Gaussian noise (WGN) in many scenarios. However, there is little literature on the channel capacity under the mixed noise. In this paper, we analyze the existence of the capacity under p-th moment constraint and prove that there are only finite mass points in capacity-achieving distribution. Furthermore, we discuss the capacity relating to specific modulation format under the mixed noise and derive a lower bound with closed form. Numerical results reveal that the capacity of the mixed noise channel will decrease as the impulsiveness of the mixed noise becomes significant and the obtained analytical capacity lower bound is tight.

In many practical applications, the communication system suffers from mixed Gaussian-impulsive noise consisting of both Gaussian white noise and non-Gaussian impulsive noise. Obtaining the optimal signal detection algorithm under these scenarios requires the estimate of the mixed Gaussian-impulsive noise parameters. Unfortunately, the estimation accuracy will deteriorate in the presence of the transmitted signal. To solve the problem, we propose a blind source separation method with a neural network, namely U-net++, to separate the transmitted signal from the mixed noise. Then, the parameter estimation can be implemented with the recovered noise samples. An adaptive clipping preprocessing module and a novel loss function of the network are designed based on the statistical property of the mixed noise. Results show that our algorithm outperforms existing baselines on both source separation and parameter estimation.