KBKostas Berberidis

Kostas Berberidis

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Kostas Berberidis has a total of 43 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Improving Wideband Massive MIMO Channel Estimation With UAV State-Space Information

    by Evangelos Vlachos, Christos Mavrokefalidis, Kostas Berberidis, George C. Alexandropoulos

    In this paper, we focus on the estimation of massive Multiple-Input Multiple-Output (mMIMO) channels in high-frequency point-to-point wireless communications between a terrestrial multi-antenna base station, realizing fully digital beamforming, and a multi-antenna Unmanned Aerial Vehicle (UAV), capable of only analog combining via a single reception Radio-Frequency (RF) chain. A wideband channel model, incorporating selectivity which is usually referred to as beam squint and is particularly relevant in high frequency bands, such as millimeter waves and terahertz, is considered. To account for the UAV's limited capability to collect long training sequences, due to its single-RF reception architecture, and the beam-squint effect at both communication ends, we present a novel channel estimation approach that exploits information provided by the control module of the UAV, namely its state-space vector. The proposed approach comprises a hybrid parametric and non-parametric estimation optimization framework that is efficiently solved through an iterative algorithm using the alternating direction method of multipliers. Our extensive performance investigations, including comparisons with benchmark schemes, showcase that considerable gains can be achieved if state-space information from the UAV's control module is appropriately incorporated in the channel estimation process.


  • Open Access

    A Pre-Activation Residual Convolutional Network With Attention Modules for High-Resolution Segmented EEG Emotion Recognition

    by Ioannis Charalampous, Christos Mavrokefalidis, Kostas Berberidis

    Emotion recognition based on electroencephalography (EEG) signals has attracted considerable research interest over the past few years and several potential applications have been proposed such as enhancing human-computer interaction, improving mental health diagnosis, and fine-tuning the customer experience at the marketing level. This paper introduces a novel model, called Pre-Activation Residual Convolutional Network with Attention Modules (PRCN-AM), designed to enhance the accuracy and robustness of emotion recognition based on EEG signals. PRCN-AM combines residual convolutional layers with pre-activation and attention modules to effectively capture and analyze the complex spatial-temporal patterns inherent in EEG signals. Two experimental procedures, namely, subject-dependent and subject-independent, were conducted and different time segmentations on the preprocessing stage were tested. The suggested exploitation of the temporal dynamics of the EEG signals in emotion recognition turns out to be useful, as classification accuracies of up to 99.51% and 97.51% on SEED and SEED-IV datasets have been achieved, respectively, thus, outperforming the current state-of-the-art models.


  • Extending Online 4D Situational Awareness in Connected and Automated Vehicles

    by Nikos Piperigkos, Christos Anagnostopoulos, Aris S. Lalos, Kostas Berberidis

    Empowering each vehicle with four dimensional (4D) situational awareness, i.e., accurate knowledge of neighboring vehicles' 3D locations over time in a cooperative manner (instead of focusing only on self-localization), is fundamental for improving autonomous driving performance in diverse traffic conditions. For this task, identification, localization and tracking of nearby road users is critical for enhancing safety, motion planning and energy consumption of automated vehicles. Advanced perception sensors as well as communication abilities, enable the close collaboration of moving vehicles and other road users, and significantly increase the positioning accuracy via multi-modal sensor fusion. The challenge here is to actually match the extracted measurements from perception sensors with the correct vehicle ID, through data association. In this paper, two novel and distributed Cooperative Localization or Awareness algorithms are formulated, based on linear least-squares minimization and the celebrated Kalman Filter. They both aim to improve ego vehicle's 4D situational awareness, so as to be fully location aware of its surrounding and not just its own position. For that purpose, ego vehicle forms a star like topology with its neighbors, and fuses four types of multi-modal inter-vehicular measurements (position, distance, azimuth and inclination angle) via the linear Graph Laplacian operator and geometry capturing differential coordinates. Moreover, a data association strategy has been integrated to the algorithms as part of the identification process, which is shown to be much more beneficial than traditional Hungarian algorithm. An extensive experimental study has been conducted in CARLA, SUMO and Artery simulators, highlighting the benefits of the proposed methods in a variety of experimental scenarios, and verifying increased situational awareness ability. The proposed distributed approaches offer high positioning accuracy, outperforming other state-of-the-art centralized and distributed methods.


  • Open Access

    Calendar [Dates Ahead]

    by Samantha Walter, Petros Maragos, Kostas Berberidis, Petros T. Boufounos, Maria Torres, Katrien De Moor

  • Connections Between Deep Equilibrium and Sparse Representation Models With Application to Hyperspectral Image Denoising

    by Alexandros Gkillas, Dimitris Ampeliotis, Kostas Berberidis

    In this study, the problem of computing a sparse representation of multi-dimensional visual data is considered. In general, such data e.g., hyperspectral images, color images or video data consists of signals that exhibit strong local dependencies. A new computationally efficient sparse coding optimization problem is derived by employing regularization terms that are adapted to the properties of the signals of interest. Exploiting the merits of the learnable regularization techniques, a neural network is employed to act as structure prior and reveal the underlying signal dependencies. To solve the optimization problem Deep unrolling and Deep equilibrium based algorithms are developed, forming highly interpretable and concise deep-learning-based architectures, that process the input dataset in a block-by-block fashion. Extensive simulation results, in the context of hyperspectral image denoising, are provided, which demonstrate that the proposed algorithms outperform significantly other sparse coding approaches and exhibit superior performance against recent state-of-the-art deep-learning-based denoising models. In a wider perspective, our work provides a unique bridge between a classic approach, that is the sparse representation theory, and modern representation tools that are based on deep learning modeling.


  • Open Access

    Efficient Distributed Multi-Task Schemes for mmWave MIMO Channel Estimation

    by Maria Trigka, Christos Mavrokefalidis, Kostas Berberidis

    In this study, the problem of sparse channel estimation is investigated with the employment of a fully distributed approach. We exploit the spatially joint sparsity structure of the involved channels to formulate the channel estimation problem in the angular domain. The devices collaboratively estimate their channel sparsity support sets before the local estimation of the channel values, assuming the existence of global and common support subsets. The combination of the proposed distributed scheme with the classical Simultaneous Orthogonal Matching Pursuit (SOMP) algorithm is calledWeighted Distributed Simultaneous Orthogonal Matching Pursuit (WDiSOMP). The performance of WDiSOMP is assessed under a multi-task scheme considering a new weighted voting method. The new mechanism is applied at each support set index and enhances its estimation accuracy and, thus, the eventual estimation of the overall channel. The mean squared error (MSE) is utilized to derive the performance bounds and assess the efficacy of the WDiSOMP estimator. Finally, the performance and the theoretical findings are evaluated via the comparison of WDiSOMP with Distributed Simultaneous Orthogonal Matching Pursuit (DiSOMP), local SOMP, and a centralized approach based on Structured SOMP (SSOMP) in terms of the MSE.


  • Optimal MmWave Sensor Selection for Bearing-Only Localization in Smart Environments

    by Evangelos Vlachos, Evangelos D. Spyrou, Chrysostomos Stylios, Kostas Berberidis

    Nowdays, millimeter wave (mmWave) direction sensors are being used increasingly as general-purpose radars, since they can provide high-level of accuracy for a variety of situations at low-cost. Via mutliple mmWave sensors, bearing estimation can be derived to track the position of a target, while in smart environments several sensors can be deployed. In this work, we provide an optimal sensor selection technique, for choosing which sensors to activate for bearing estimation and which not. The proposed approach is decomposed into training phase, where sensor selection is performed, and operational phase, where bearing estimation is obtained. Via simulation results we evaluate the proposed approach compared with the conventional methodology of utilizing all available data streams.


  • Graph Laplacian Diffusion Localization of Connected and Automated Vehicles

    by Nikos Piperigkos, Aris S. Lalos, Kostas Berberidis

    In this paper, we design distributed multi-modal localization approaches for Connected and Automated vehicles. We utilize information diffusion on graphs formed by moving vehicles, based on Adapt-then-Combine strategies coupled with the Least-Mean-Squares and the Conjugate Gradient algorithms. We treat the vehicular network as an undirected graph, where vehicles communicate with each other by means of Vehicle-to-Vehicle communication protocols. Connected vehicles perform cooperative fusion of different measurement modalities, including location and range measurements, in order to estimate both their positions and the positions of all other networked vehicles, by interacting only with their local neighborhood. The trajectories of vehicles were generated either by a well-known kinematic model, or by using the CARLA autonomous driving simulator. The proposed distributed and diffusion localization schemes significantly reduced the GPS error and do not only converged to the global solution, but they even outperformed it. Extensive simulation studies highlight the benefits of the various methods, which in turn outperform other state of the art approaches. The impact of the network connections and the network latency are also investigated.


  • Open Access

    Coalition Formation Games for Improved Cell-Edge User Service in Downlink NOMA and MU-MIMO Small Cell Systems

    by Panagiotis Georgakopoulos, Tafseer Akhtar, Christos Mavrokefalidis, Ilias Politis, Kostas Berberidis, Stavros Koulouridis

    In today’s wireless communication systems, the exponentially growing needs of mobile users require the combination of new and existing techniques to meet the demands for reliable and high-quality service provision. This may not always be possible, as the resources in wireless telecommunication systems are limited and a significant number of users, usually located at the cell edges, can suffer from severe interference. For this purpose, a new Joint Transmission Coordinated Multipoint (JT-CoMP) scheme, in which the transmission points’ clustering is based on a coalition formation game, is deployed alongside with Non-Orthogonal Multiple Access (NOMA) in a Cloud Radio Access Network (C-RAN) consisting of small cells. To further enhance the network’s performance, Multiuser Multiple-Input Multiple-Output (MU-MIMO) with Zero-Forcing (ZF) beamforming is applied. The proposed scheme’s performance, in terms of user throughput, is then compared to that of a scheme where JT-CoMP with static clustering is selected, a JT-CoMP scheme with clustering based on a greedy algorithm and a scenario where JT-CoMP is not deployed. Simulation results confirm that the proposed scheme reliably improves the throughput of users with poor wireless channels while guaranteeing that the performance of the rest is not severely undermined.