ABSTRACT Content‐Based Image Retrieval (CBIR) systems have difficulties with computing efficiency, illumination robustness and noise sensitivity. Traditional methods rely on handcrafted features or monolithic deep learning architectures, which either lack adaptability to diverse image domains or suffer from high computational complexity. To bridge this gap, a unique two‐tier deep learning system is presented in this research to overcome these drawbacks. First, a supervised neural network (SNN) reduces dimensionality and improves interpretability by converting HSV colour space into semantic 2D colour labels through pixel‐level classification. This addresses the inefficiency of processing raw RGB data while preserving illumination‐invariant colour semantics. Second, a Convolutional Neural Network (CNN) greatly increases computing efficiency by processing these labels rather than raw images. By operating on compressed 2D representations, the system achieves faster inference compared to standard 3D CNN pipelines. The framework presents Variable Weight Overall Similarity (VWOS), a versatile similarity metric that combines semantic (softmax) and structural (MaxPool3) elements with dynamically predicted weights using a neural network to automatically optimise retrieval performance based on image content. This adaptive fusion resolves the limitations of fixed‐weight similarity measures in handling heterogeneous query types. The system has achieved a performance with precision@10 scores of 0.9‐1.0 and classification accuracies of 0.85‐0.98 when tested on the PH 2 , Oxford Flowers, Corel‐1k, Caltech‐101 and Kvasir datasets. Notably, it outperforms current handcrafted, deep learning and hybrid approaches, achieving 1.0 precision@10 on four datasets and 0.96 accuracy on medical Kvasir images. Quantitative comparisons show 9%–14% higher precision than handcrafted methods, 3%–35% improvement over deep learning baselines, and 12% better than hybrid systems. This approach is especially promising for applications involving multimedia retrieval and medical imaging, where interpretability and accuracy are crucial.

We study the secrecy sum-rate (SSR) of a multi-user system in the presence of artificial noise (AN) and determine conditions under which it is necessary to add AN assuming eavesdropper (Eve) cannot remove the interference. We first derive a closed-form expression for the asymptotic SSR of the system and determine the optimum power allocation. The closed form expressions for the SSR and optimum power allocation closely match our simulation results. We also show that as the number of receive antennas at Eve increases, we need to allocate more power to AN in order to degrade reception at Eve.

We derive the asymptotic ergodic secrecy sum rate (ESSR) of block diagonalization precoding for physical-layer security in downlink multi-user wireless networks that use artificial noise. First, we derive the ESSR of the system. Then, we derive the power allocation and the asymptotic ESSR as the number of antennas increases. After, we derive an approximate expression for the optimum ratio of the number of transmit antennas to the number of receive antennas (Ψ). We then introduce the user dropping factor (UDF) as a means to derive the optimum number of users in the system under different scenarios and then derive the asymptotic optimum UDF, which gives the optimum number of scheduled users. We also study the effect of imperfect channel state information and derive a closed-form ESSR. Simulation results show that these precoders outperform channel inversion (CI) and regularized CI schemes in terms of the ESSR at all SNRs. We also show that as the number of receive antennas of each user increases, the asymptotic ESSR gap also increases.

We analyze asymptotic performance of regularized block diagonalization (RBD) precoding along with artificial noise (AN) used for physical layer security in downlink multi-user wireless networks. We derive the secrecy and asymptotic secrecy sum-rates for this scheme. We also derive closed form expressions for the asymptotic power allocation to the AN signal and legitimate users. Simulations show RBD precoding outperform regularized channel inversion (RCI) in terms of secrecy sum-rate with a margin of 0.5 bits/s/Hz.

We derive the asymptotic secrecy sum-rate of block diagonalization (BD) precoding for physical layer security in downlink multi-user wireless networks using artificial noise (AN). We first derive the secrecy sum-rate assuming that the channel of the eavesdropper is not known at the transmitter and then the asymptotic secrecy sum-rate is derived. The optimum power allocation for asymptotic secrecy sum-rate is derived after that. Simulation results show that the proposed precoders outperform channel inversion (CI) and regularized channel inversion (RCI) schemes in terms of the secrecy sum-rate at all SNRs. We also show that as the number of receive antennas of each user increases, the sum-rate gap between the proposed scheme and existing methods increases.

Abstract We propose two novel clustered linear precoding schemes applicable to network multi‐input multi‐output systems using only partial channel state information to enhance the sum‐rate of the system. Using a channel model that decomposes a multi‐input multi‐output channel matrix into transmit and receive steering vectors and assuming that only transmit steering vectors are available at the base transmit stations, we, first, propose a regularized channel inversion precoding scheme to enhance the sum‐rate assuming only single‐antenna users are available in the system. Next, because of the limitation of regularized channel inversion to handle users with multiple receive antennas, a novel block diagonalization method is proposed. We construct the precoding matrices that jointly eliminate inter‐cell interference and maximize the sum‐rate for a given input covariance matrix. Assuming total power constraint and per‐base‐station power constraints, optimal power allocation schemes are further developed to optimize the sum‐rate. We analytically show that the sum‐rate increases linearly with the number of users when only single‐antenna users are present in the system. Numerical results show that at low signal‐to‐noise ratios, the block diagonalization precoding outperforms the regularized channel inversion in terms of the bit error rate; while at high signal‐to‐noise ratios, the regularized channel inversion provides a better performance. Copyright © 2016 John Wiley & Sons, Ltd.

In this paper, the problem of precoder design and node power allocation in a three node cooperative multiple-input multiple-output (MIMO) relay system is discussed. All the nodes are assumed to be equipped with multiple antennas and the source and the relay nodes are subject to a total transmit power constraint. We first introduce an upper bound on the received mean square error (MSE), then we suggest suboptimal structures to the source and the relay precoders to minimize this bound. Unlike the previous works in this area, the proposed precoders are easy to calculate and does not involve in computationally costly iterative methods. Next, based on the precoder structures, it is shown that proper transmit power assignment to the source and relay nodes can further improve the system performance and the optimal power allocation for the special case of single symbol transmission is derived. Simulation results show significant improvement in the received mean square error (MSE) by applying the proposed precoding and node power allocation schemes. The node power allocation scheme is also shown to be most effective when the relay and destination nodes have different noise levels.

Multicell coordination increases the cell-edge users' throughput in cellular wireless communication if the inter-cell interference (ICI) is mitigated properly. In this paper, we use the null space concept to design precoders and decoders to remove ICI for cell-edge users in the downlink of a network MIMO system. We propose two novel methods that are able to eliminate ICI assuming no channel state information (CSI) is available at the basestations. Calculating signal-to-interference-ratio (SINR), we show that our proposed methods reaches to higher SINR in comparison to the block diagonalization (BD) with full CSI. By simulations, we show our methods outperform the BD method with full CSI in terms of BER since the SINR is higher for our proposed methods. We also show that to obtain most of sum-rate provided by coordination strategy, it is not necessary to have global coordination and having four cells coordinating with each other is enough to reach more than 85 percent of maximum achievable sum-rate.