Several components in the design of next-generation networks, including user profiling and network slicing, rely on accurate models of traffic load. In this context, recent studies have focused on various video traffic categories, while traffic associated with extended reality (XR) services has received limited attention. This paper introduces a novel empirical model for 3D XR traffic, developed by encoding real Point Clouds using a standard-compliant codec, and able to account for the dynamic of service sessions and user behaviors over an entire session. Our methodology encompasses multiple temporal scales, ranging from milliseconds to minutes, to account for different phenomena related to both user behaviour and encoder settings. Initially, we investigate the packet size distribution at the time scale of a semantic unit, corresponding to the encoding of a single point cloud. We verify that it can be effectively represented by a heavy-tailed Gamma distribution. Then, we illustrate how this insight can be leveraged to model application-layer phenomena. Specifically, we demonstrate the applicability of a general semi-hidden Markov model to capture both the temporal dynamics of service sessions and user behaviors. We provide results in terms of comparison of the empirical and fitting traffic distributions, based on quantile to quantile analysis and statistical tests. We also show how the model can be trained on real data and we provide a pseudo-code demonstrating the model application within a network simulator.

Real-world networks are typically described in terms of nodes, links, and communities, having signal values often associated with them. The aim of this letter is to introduce a novel Compound Markov random field model (Compound MRF, or CMRF) for signals defined over graphs, encompassing jointly signal values at nodes, edge weights, and community labels. The proposed CMRF generalizes Markovian models previously proposed in the literature, since it accounts for different kinds of interactions between communities and signal smoothness constraints. Finally, the proposed approach is applied to (joint) graph learning and signal recovery. Numerical results on synthetic and real data illustrate the competitive performance of our method with respect to other state-of-the-art approaches.

A novel cooperative least mean square adaptation rule is proposed with application to fractionally spaced blind channel equalization schemes used in quadrature amplitude modulation data links. A novel cost function based on second-order statistics only of the received signal is defined and jointly used with another chosen among classical blind cost functions such as decision directed and other Bussgang-type ones. Numerical results show that severe misconvergence observed in classical blind equalization algorithms is drastically mitigated using the here described technique.

The development of wireless personal area networks is one of the key challenges in the new frontier of communications. The standardization activities of the IEEE802.15.6 and IEEE802.15.4f workgroups along with the proposed 368-369 ECMA specifications open new roads to optimize performance in ultrawideband (UWB) communication links and services. In this perspective, new communications require high reliability, high throughput for multimedia content delivery, very low latency, and bit error rates are required. In this regard, this contribution deals with a turbo-like mechanism that, starting from an interference acquisition phase, proceeds with data-aided channel estimation and synchronization to data detection of ultrawideband impulse-radio (UWB-IR) terminals operating over broadband channels affected by multipath fading and in the presence of imperfect orthogonal access. The performance of the proposed transceiver has been evaluated in terms of both required number of pilot symbols for the channel estimation/data synchronization and bit error rate resulting from the channel estimation and interference analysis. Moreover, the performance related to system efficiency (i.e., data transmitted with respect to the whole information needed for estimation purposes) is evaluated by showing the balance between achieved transmission rate and bit error rate. Numerical comparisons have been performed as well as tests on real data and partial implementation of the proposed scheme.

This paper addresses the problem of texture images recovery from compressively sampled measurements. Texture images hardly present a sparse, or even compressible, representation in transformed domains (e.g. wavelet) and are therefore difficult to deal with in the Compressive Sampling (CS) framework. Herein, we resort to the recently defined Graph-based transform (GBT), formerly introduced for depth map coding, as a sparsifying transform for classes of textures sharing the similar spatial patterns. Since GBT proves to be a good candidate for compact representation of some classes of texture, we leverage it for CS texture recovery. To this aim, we resort to a modified version of a state-of-the-art recovery algorithm to reconstruct the texture representation in the GBT domain. Numerical simulation results show that this approach outperforms state-of-the-art CS recovery algorithms on texture images.

In compressive sensing, the Restricted Isometry Property is an analytical condition on the measurement matrix that assures reconstruction of a signal which is sparse either in the spatial or in a transformed domain given an undersampled measurements' set. In this paper, we demonstrate the RIP for a sparse, structured measurements matrix, referred to as Radon-like CS matrix. The sparse Radon-Like CS matrix favorably applies to real sensing problems, since it significantly reduces the energy/bandwidth cost of actually collecting each and every sensed values contributing to the CS measurements. Simulation results confirm the feasibility of field reconstruction from undersampled CS measurements set obtained using the Radon-Like CS matrix.

We introduce a novel, both gain and signal-to-noise ratio independent, constellation unaware, blind frequency offset estimation procedure for QAM signals. Asymptotic performance analysis and numerical simulations show that the herein presented method outperforms a selected state of the art blind constellation unaware estimator, especially for cross constellations.

In this letter, we address the problem of estimating a parameter acting as a scale factor on the observations probability density function (pdf), i.e. a scale parameter. Histogram based Maximum Likelihood (ML) estimation of a scale parameter requires the evaluation of a discrete scale correlation. We show how ML estimation can be implemented by means of a computationally efficient Discrete Fourier Transform based procedure, when geometric histogram sampling is adopted. As a case study, we analyze a gain estimator for general QAM constellations. Simulation results and theoretical performance analysis show that the presented ML estimator outperforms selected state of the art estimators, approaching the Cramér-Rao Lower Bound (CRLB) for a wide range of SNR.

In this paper we develop a strategy for visual relevancy evaluation using Rate Distortion analysis in the Circular Harmonic Functions (CHF) Domain. In the CHF domain, the different visual characteristics of the transformed frames - edges, lines, forks, crosses, etc - are emphasized. Resorting to this domain we analyze a new approach to evaluate the distortion due to an event of video data loss accounting for the visual relevance of individual lost data. In a real video streaming session, distortion measurements taking into account the visual impact of error events can be stored along with pre-coded video sequence so as to drive the strategies for ad-hoc error protection. Numerical simulations show that the approach herein proposed to evaluate the visual impact of loss events in the decoded sequences captures the visual distortion much better than commonly adopted mean square error criteria between the decoded and the original sequence.

In this paper we present theoretical performance analysis for a blind frequency offset estimator for cross quadrature amplitude modulated constellations. The estimator is based on applying a tentative frequency offset compensation by means of a nonlinear transformation of the received signal samples and on estimating an accumulation function in different angular windows. For perfect frequency offset compensation, the measurements are suitably clustered and their accumulation, named "constellation phase signature" (CPS), is a peaked function of the window orientation. Hence, the frequency offset estimator is selected by maximization of the peakness of the accumulation function. The performance analysis is shown to match the numerical simulations for medium to high values of SNR.

Fast and bit-saving video bit-rate switching is an important issue in video streaming systems on a time varying channel as the one offered by a wireless mesh network, or the one sensed during a vertical handover. The recent H.264 video coding standard supports the seamless switching among bitstreams coded at different bit-rates by means of suitably coded frames, named switching pictures. This work addresses the modelling of the traffic generated by a H.264 source performing bit-rate switching using SP frames. The H.264 source is modelled by a Markov chain where each state models the generation of an entire group of pictures (GOP), and is characterized by the kind of SP frame encoded in the GOP. Inter- frame correlation, typical of video sources, is suitably taken into account by the interstate dependence. The accuracy of the model is assessed by comparison of the cell loss rate of a fixed size buffer filled with a synthetic source according to the model herein proposed, with a state of the art AR model and with a real H.264 video codec.

In this work1we discuss a game theoretic approach to bitstream switching in video coding. Fast and bit-saving video bitstream switching is an important issue in video communication system on time varying channels. The most recent video coding standard, namely H.264, support the seamless switching among bitstreams coded at different bitrates by means of suitably coded frames, named Switching Pictures. Since the rate-distortion characteristics of switching frames differ from those of I and P frames, their location affect both the bit-rate and the quality of the coded sequence. In this work, we address the optimization of the SP frames location under an assigned bitrate budget. At this aim we restor to a game theoretic approach and we show that the optimal solution is met when the SP frames are assigned to the frame with the smallest innovation. Experimental results show the advantage in terms of both rate and distortion achieved by the optimized Switching frame insertion with respect to basic H264 coding.

We introduce a new blind phase offset estimator for general quadrature amplitude modulated (QAM) signals. The estimator is based on the computation of a suitable phase distribution that we call "signature". The signature is defined as the phase-dependent distribution of the received signal magnitude after the application of a nonlinear transformation. The signature of a QAM signal is constituted by a discrete number of pulses and it has good autocorrelation properties in the sense of maximum/side-lobe ratio. Since the effect of a phase offset is a cyclic shift of the signature, the phase offset can be estimated by searching for the maximum of the cyclic cross-correlation between the zero-phase signature of the expected constellation, and the signature calculated on the received signal. The resulting estimator is characterized by a low computational complexity and does not need gain control. The comparison shows that the presented estimator is asymptotical efficient and outperforms existing estimators for medium to high values of SNR, especially for complex constellations

In this paper, we present a texture classification procedure that makes use of a blind deconvolution approach. Specifically, the texture is modeled as the output of a linear system driven by a binary excitation. We show that features computed from one-dimensional slices extracted from the two-dimensional autocorrelation function (ACF) of the binary excitation allows representing the texture for rotation-invariant classification purposes. The two-dimensional classification problem is thus reconduced to a more simple one-dimensional one, which leads to a significant reduction of the classification procedure computational complexity.

In this contribution, we show how a hierarchical edge image analysis at macroblock, block and subblock levels allows identifying preminent directional image components. The analysis, performed in the Radon domain, selects the scale level at which the image features present a directional structure that can be better reconstructed by a directional spatial interpolation. When the directional information is known at the decoder side it can drive the spatial error concealment. The experimental results, referring to the case of H.264 coded video, show the significant improvement of the decoded video visual quality achievable by the described technique.

We discuss different architectures for error resilient video streaming. The architectures exploit a resilient video coding scheme, based on out-of-band transmission of suitable redundant information, that counteracts synchronization loss between the decoder and the bitstream due to transmission errors. We investigate unicast and multicast video streaming solutions and show a power/bandwidth efficient hybrid unicast/multicast architecture. Experimental results show the decoded video quality obtained by adopting the resilient streaming architecture in the case of H.264 coding.

This work describes how the introduction of redundancy by means of coding constraints known at the decoder side improves the error resilience of a video communication scheme. The decoder can exploit the a priori knowledge of the constraints to verify the integrity of the decoded bitstream. If the transmission errors cause the violation of the coding constraints, the decoder detects an error and recovers the synchronization with the bitstream. The improved detection capability results in better localization of the decoded frame areas affected by transmission errors, so allowing more accurate error concealment. The adoption of coding constraints does not require any feedback channel between the decoder and the coder since the decoder can be notified of the constraints during the initial negotiation phase. The syntax of the coder output is not affected and the coded bitstream remains standard compliant. Therefore, this technique is quite general and it is well suited to many transport schemes, either packet switched or circuit switched. Furthermore, it can be easily integrated with different error resilience tools. The improvement achievable in the decoded image quality using the coding constraints is here shown in the case of H.264 coded video.

In this paper, a model based texture classification procedure is presented. The texture is modeled as the output of a linear system driven by a binary image. This latter retains the morphological characteristics of the texture and it is specified by its spatial autocorrelation function (ACF). We show that features extracted from the ACF of the binary excitation suffice to represent the texture for classification purposes. Specifically, we employ a moment invariants based technique to classify the ACF. The resulting proposed classification procedure is thus inherently rotation invariant. Moreover, it is robust with respect to additive noise. Experimental results show that this approach allows obtaining high correct rotation-invariant classification rates while containing the size of the feature space.

This work addresses the multichannel Bussgang restoration algorithm for blind image deconvolution problems. The case of spatial correlated nonGaussian images is considered, and the structure of nonlinear minimum mean square error estimators, to be used in the Bussgang deconvolution algorithm, is discussed. In particular, it is conjectured that, for nonGaussian, spatially correlated random fields, a better nonlinear estimator is obtained in a suitable wavelet transformed domain. The wavelet transform here considered is the so-called circular harmonic wavelet (CHW), because of its characteristic property of capturing specific image features, such as edges, lines, forks, crosses, and so on. Experimental results pertaining to restoration of motion blurred natural images are also reported.

This work extends the Bussgang blind equalization algorithm to the multichannel case with application to image deconvolution problems. We address the restoration of images with poor spatial correlation as well as strongly correlated (natural) images. The spatial nonlinearity employed in the final estimation step of the Bussgang algorithm is developed according to the minimum mean square error criterion in the case of spatially uncorrelated images. For spatially correlated images, the nonlinearity design is rather conducted using a particular wavelet decomposition that, detecting lines, edges, and higher order structures, carries out a task analogous to those of the (preattentive) stage of the human visual system. Experimental results pertaining to restoration of motion blurred text images, out-of-focus spiky images, and blurred natural images are reported.

In this paper a model based texture classification procedure robust to sample rotation is presented. The texture is modeled as the output of a linear system driven by a binary image. This latter retains the morphological characteristics of the texture and it is specified by its spatial autocorrelation function (ACF). We show that features extracted from the ACF of the binary excitation suffice to represent the texture for classification purposes. Specifically, we employ a classical moment invariants based technique to classify the ACF and the resulting classification procedure is thus inherently rotation invariant. Experimental results show that this approach allows obtaining high correct rotation-invariant classification rates while reducing the size of the feature space and the computational burden.

In this paper, an unsupervised model-based texture reproduction technique is described. In accordance with the Julesz's (1962) conjecture, the statistical properties of the prototype up to the second order are copied in order to generate a synthetic texture perceptually indistinguishable from the given sample. However, this task is accomplished using a hybrid approach which operates partially in the spatial domain and partially in a multiresolution domain. The latter employed is the circular harmonic function (CHF) domain since it has been proven to be well suited for mimicking the behavior of the human visual system (HVS). This approach allows, for a wide range of textures typologies, obtaining synthetic textures that better match the prototype with respect to the ones obtained using techniques based on the Julesz's conjecture operating only in the spatial domain, and to dramatically reduce the computational complexity of similar methods operating only in the multiresolution domain.

The problem of image texture stochastic modeling by means of the 2D Wold decomposition is addressed. Specifically, the 2D Wold decomposition additively separates a texture into a deterministic and an indeterministic component. Since these components are shown to be statistically orthogonal, for textures having a distribution that significantly deviates from Gaussianity, it is possible, in principle, to establish some link between the two components. The aim of this contribution is to show how the indeterministic component can be predicted from the deterministic one by means of suitable nonlinear schemes. The hierarchical structure of the 2D Wold decomposition is thus generalized to deal with the non-Gaussian behavior of most of the natural textures of interest.

An unsupervised color texture synthesis-by analysis method is described. The texture is reproduced to appear perceptually similar to a given prototype by copying its statistical properties up to the second order. The synthesized texture is obtained at the output of a single-input three-output nonlinear system driven by a realization of a white Gaussian random field. Significant complexity reduction is gained by exploiting the rank deficiency of the cross power spectral density matrix of the color texture samples.

This contribution describes a fast frequency domain approach for blind channel identification which does not rely on the statistics of the symbols. The proposed approach is based on the so-called "intraspectral relations" of DFTs of PAM fractionally sampled signals. The use of DFTs is allowed under certain conditions commonly encountered in data communication systems. The intraspectral relations are equivalent to analogous relations introduced in the time domain in Xu et al. (1995), Moulines et al. (1995) and Hua (1996). From the intraspectral relations, asymptotically efficient solutions are derived which turn out to be either more accurate or less expensive in terms of complexity w.r.t. the time domain counterparts. Simulation results are provided to assess the validity of the proposed approach in comparison with the Rao Cramer bound and with other approaches from the literature.

We address the problem of the blind joint identification and order estimation of a non-minimum phase FIR communication channel by exploiting the cyclo-stationarity of the received signal sampled at a rate greater than the symbol rate. We show that the identification can be formulated as a "subspace fitting" problem; this allows for using the subspace distance as a test statistic to detect the correct channel length (among different hypotheses). Moreover, mimicking estimation procedures proposed in the framework of DOA estimation, an asymptotically efficient procedure is proposed which obtains the channel estimate in two steps: the channel estimate obtained in the first step determines optimal weights which are then employed in the second step to refine the channel estimate. The accuracy of the identification is comparable with other methods described in the literature (e.g. the subspace method) while the test for order detection performs quite well also in the presence of channel disparity outperforming commonly used tests based on the eigenvalues of the covariance matrix.

An innovative unsupervised color texture, model based, synthesis method is presented. It aims to reproduce a color texture perceptually similar to a given prototype. To achieve this task, we reconstruct the texture on a stochastic base, matching the prototype's statistics up to the second order to the synthesized texture ones. The model employed allows to replace the multiple input multiple output filters, which have to be used in the general case, with more simple structures, leading to a dramatical computational complexity reduction without affecting the results quality.

A twin stage texture synthesis-by-analysis method is presented. It aims to approximate first- and second-order distributions of the texture, accordingly to the Julesz conjecture. In the first stage, the binary textural behavior of a given prototype is represented by means of a hard-limited Gaussian process. In the second stage, the texture is synthesized by passing the binary hard-limited Gaussian process through a linear filter followed by a zero memory histogram equalizer.

A new technique for unsupervised texture synthesis by analysis is presented. It employs a stochastic model of textured fields based on white Gaussian fields passed through a nonlinear system composed of two linear filters connected by a hard limiter. The identification of the texture models is performed using a Bussgang blind deconvolution algorithm and exploiting of the Van Vleck rule for determining the filters parameters. After a theoretical discussion of the method typical examples are provided.

The polarity coincidence estimator of the spectral centroid counts the net number of times the phasor of a stationary Gaussian process crosses a chosen phase-reference straight line through the origin during a time T. By following Pawula's (1968) approach, the covariance of two different simultaneous polarity-coincidence estimates having distinct phase reference lines is derived. This basic result is applied to compute the variance of a generalized polarity-coincidence estimator considering N arbitrarily chosen reference lines.>

The extension of higher order moments based spectral analysis to hybrid nonlinear (HNL) moments is theoretically discussed. It is shown that the structure of the covariance matrix of harmonic processes is retained by HNL moments matrices, thus allowing the use of eigenassisted estimation procedures. The Gaussian components insensitivity of slice cumulants is generalized to a subclass of HNL moments, and a covariance subtraction method for blind rejection of colored noise is proposed. Simulation results confirm the applicability of such a procedure for bearings estimation in array processing schemes.>

The complementary aspects of spectrum estimation based on higher-order order statistics are considered. Higher-order analysis of stochastic array data is formalized in a general fashion through the introduction of the class of hybrid nonlinear (HNL) moments. HNL moments comprise, as a special case, slice cumulants and their mixtures. In this framework, the case of joint use of covariance and fourth-order slice cumulant is handled in a simple and direct way. Simulation results confirm the theoretical analysis.>

A novel, suboptimal, nonlinear recursive algorithm for the estimation of filtered Markov chains from noisy observations is presented. The expression of the estimator is formally identical to the Kalman filter solution when the error covariance matrix is replaced by the conditional covariance matrix. Starting from the Martingale-difference (MD) representation theorem, a recursive expression for the conditional covariance matrix that can be reasonably approximated by means of the Markov property is derived. Thus the obtained estimator is suboptimal in the sense that the conditional covariance matrix is evaluated only approximately.>

The function e-x2has many well-known mathematical properties. Here it is shown that if a stationary process has a Gaussian-shaped acf, then its PARCOR coefficients decay exponentially.