engleski

Multiple-Threshold Estimators for Impulsive Noise Suppression in Multicarrier Communications

The performance of digital communication systems, employed in cellular, broadcasting, and wireless access networks, can be significantly degraded by adverse channel conditions and interferences, which characterize wireless communications in urban environments. These interferences are typically modeled as non-Gaussian impulsive noises [1]–[3], such as the Middleton's Class-A noise that, together with the associated suppression techniques, has been widely investigated in the past (see [4]–[8] and references therein). Although multicarrier modulations, currently employed in most of the wireless communication systems, are inherently more resistant to impulsive noise (ImpN) than single carrier modulations, the counteraction of the performance degradation caused by ImpN is still a challenging research area for communication engineers [9]–[11]. Actually, countermeasures for Class-A noises can be easily generalized to any scenario characterized by multi- component Gaussian mixture noises [7], [8], and consequently also to alpha-stable impulsive noises, which can be modeled as a Gaussian mixture as well [12]– [14]. Assuming a Gaussian source impaired by memoryless ImpN, optimum system performance in terms of mean-squared error (MSE) and signal-to-noise power ratio (SNR) can be achieved by applying a Bayesian signal estimator [15]. Specifically, the optimal Bayesian estimator (OBE) for real-valued Gaussian mixture noise has been proposed in [8], and successively extended to complex signals in [9]. Although the OBE guarantees optimal MSE and SNR for uncorrelated ImpN, in some cases its implementation in practical receivers may be cumbersome or not attractive. For instance, if ImpN suppression is performed before A/D conversion, an analogic solution may be too complex, while if impulse noise suppression is performed after A/D conversion, the computational complexity and the system latency may be relatively high, especially for simple low-cost devices [8], [15]. Consequently, less complex solutions are typically based on signal thresholding, associated with blanking (nulling) [6], [8], [10], [16], clipping [7], [8], [15], or their combinations [5], [7], [11], [17]. Recently, an MMSE estimator, constrained to a given quantization resolution of the noisy observations [15], has been proved to converge to the MMSE optimal solution, e.g., the OBE, which conversely uses an infinite resolution of the observation. Although all the mentioned techniques can be used in any system impaired by additive ImpN, their application to OFDM, or any other multicarrier wireless system, is more challenging when the system is also affected by frequency-selective fading channels, as we will clarify.

OFDM ; Signal to noise ratio ; Fading channels ; Wireless communication ; Bayes methods ; Frequency-domain analysis ; Receivers

nije evidentirano