2025. 4^th Issue

Volume XVII, Number 4

Table of contents 

Full issue  

 

PAPERS FROM OPEN CALL

 

Margit Antal
Evaluation of Embedding Models for Hungarian Question-Answer Retrieval on Domain-Specific and Public Benchmarks
Embedding models have become a fundamental component of modern natural language processing, yet their performance in morphologically rich, low-resource languages such as Hungarian remains underexplored. In this paper, we present a systematic evaluation of state-of-the-art embedding models for Hungarian question–answer retrieval. We construct two complementary evaluation datasets: (i) a domain-specific corpus collected from company documentation, preprocessed into topical chunks with human-verified question–answer pairs and (ii) the publicly available HuRTE benchmark. Using Chroma as the vector database, we compare eight multilingual and cross-lingual embedding models alongside keyword-based search baseline. Performance is measured using Mean Reciprocal Rank (MRR) and Recall@k. Results show substantial variation across mod- els and datasets, with notable differences between domainspecific and general-purpose retrieval tasks. BGE-M3 and XLM-ROBERTA achieved the highest accuracy (MRR: 0.90) on the Clearservice dataset, while GEMINI demonstrated superior performance on HuRTE (MRR: 0.99). We complement the evaluation with comprehensive error analysis, highlighting challenges posed by Hungarian domain-specific terminology, synonyms, and overlapping topics, and discuss trade-offs in efficiency through index build time and query latency measurements. Our findings provide a comparative study of embeddingbased retrieval in Hungarian, offering practical guidance for downstream applications and setting a foundation for future research in Hungarian representation learning. The dataset and the corresponding evaluation code are publicly accessible at https://github.com/margitantal68/hungarian-embeddings.

DOI: 10.36244/ICJ.2025.4.1
Download 

 

Hsin-Ying Liang and Chuan-Bi Lin
PAPR Reduction in OTSM Systems: A Comparative Analysis of SLM Techniques with Novel Phase Matrix Designs
Orthogonal Time Sequency Multiplexing (OTSM) represents a pivotal advancement in wireless communication technology. Nevertheless, its high Peak-to-Average Power Ratio (PAPR) imposes significant constraints on its practical applications and future development. The definition of PAPR refers to the ratio of the maximum instantaneous power to the average power of a signal, and it is commonly used to assess the performance of high-power amplifiers. When PAPR values are excessively high, they reduce the efficiency of high-power amplifiers and increase the complexity of the transmission system. To mitigate this challenge, this paper explores and evaluates the efficacy of the Selective Mapping (SLM) technique for enhancing PAPR performance in OTSM systems. Leveraging the unique two-dimensional data structure inherent to OTSM, a specialized SLM approach is introduced in this paper. The proposed SLM method incorporates a Phase Generation Mechanism (PGM) that utilizes a pre-constructed perturbation phase matrix. This matrix undergoes cyclic shifts to produce multiple perturbation phase matrices. To assess the effectiveness of the proposed SLM technique, this paper investigates three distinct perturbation phase matrix generation mechanisms: Zadoff-Chu Transform (ZCT) matrices, Discrete Cosine Transform (DCT) matrices, and Randomly Generated Phase (RGP) matrices. Additionally, for evaluating PAPR performance improvement, the Complementary Cumulative Distribution Function (CCDF) is used, a statistical method that estimates the probability of high PAPR occurrences. Simulation results indicate that the RGP-based phase generation mechanism consistently outperforms the other methods in achieving significant PAPR reduction.

DOI: 10.36244/ICJ.2025.4.2
Download 

 

Yasmeen Taha Yaseeen, Mohd Fadzli Mohd Salleh, and Taha A. Elwi
Design of Enhanced Dual Band Rectenna with Binary Coding Technique of Genetic Algorithm
In this paper, an artificial intelligence-based rectenna design is proposed for Wi-Fi applications. The rectenna design is optimized using a Genetic Algorithm (GA) integrated with a Binary Coding (BC) scheme. The proposed rectenna is configured to operate at 2.45 GHz and 5.8 GHz with a maximum size of 27×30×10 mm³. The performance of the optimized rectenna has been characterized in terms of S-parameters, bandwidth, radiation patterns, and gain. For this, a dual-bandwidth patch is designed to suit the applications of 2.45 GHz and 5.8 GHz. The measured radiation patterns and S11 spectra are evaluated to obtain a peak radiation efficiency of 52% and 56% to realize a gain of 6.2 dB at 2.45 GHz and 7.12 dB at 5.8 GHz, respectively. The proposed rectenna is integrated with an RF–DC rectifier (RFD102A module) to evaluate the harvested power in terms of DC output voltage under outdoor conditions. The maximum obtained harvested DC output voltage is found to be 2.27V and 2.3V at 2.45GHz and 5.8GHz, respectively. Finally, the obtained measurements are compared to the simulated results to realize good agreements between them.

DOI: 10.36244/ICJ.2025.4.3
Download 

 

Ádám Kiss, and László Schäffer
Significant performance improvement in polarization-diversity GMSK reception using atypical filters
This paper presents novel signal processing methods to enhance the reception performance of Gaussian Minimum Shift Keying (GMSK) signals from pico- and nanosatellites, emphasizing software-defined approaches over hardware upgrades. Atypical filtering techniques, including phase-domain median and FIR filtering, as well as polarization-diverse multi-channel methods, are explored and evaluated. Real-world experiments were conducted using coherently sampled dual-polarization channels from a ground station in Szeged, Hungary, receiving transmissions from the MRC-100 satellite. Various single- and multichannel preprocessing strategies were benchmarked using packet decoding success and bit-error rates. Results show that nonlinear phase filtering and blind source separation techniques, notably FastICA, significantly increase the number of correctly decoded packets – achieving up to a 16% improvement compared to conventional demodulation without preprocessing. This study demonstrates the utility and relative independence of these methods and highlights their potential for improving satellite data throughput with no hardware modification. These techniques are suitable for integration into existing ground stations to enhance data reception performance.

DOI: 10.36244/ICJ.2025.4.4
Download 

 

Yuewen Xu
AMC-Transformer: Automatic Modulation Classification based on Enhanced Attention Model
High-accuracy automatic modulation classification (AMC) is essential for spectrum monitoring and interferenceaware access in future 6G systems [1]. We propose AMCTransformer, which tokenizes raw I/Q sequences into fixedlength patches, augments them with learnable positional embeddings, and applies multi-layer, multi-head self-attention to capture global temporal–spatial correlations without handcrafted features or convolutions. On RadioML2018.01A, our model achieves 98.8% accuracy in the high-SNR regime (SNR at least 10 dB), showing higher accuracy than a CNN and a ResNet reimplementation by 4.44% and 1.96% in relative terms; averaged across all SNRs, it also improves upon MCformer, CNN, and ResNet baselines. Consistent gains are observed on the RadioML2016.10A dataset, further validating robustness across benchmarks. Ablations on depth, patch size, and head count provide practical guidance under different SNR regimes and compute budgets. These results demonstrate the promise of transformer-based AMC for robust recognition in complex wireless environments.

DOI: 10.36244/ICJ.2025.4.5
Download 

 

Qian Hu, Zhongqiang Luo, and Wenshi Xiao
Blind Source Separation Spectrum Detection Method Based on Wavelet Transform and Singular Spectrum Analysis
To address the issue of reduced detection performance due to the impaired separation mechanism affected by noise, this paper proposes a blind source separation (BSS) detection method based on Wavelet Transform (WT) and Singular Spectrum Analysis (SSA). Firstly, the input signal is denoised using WT. Then, SSA is employed to denoise and reduce the dimension of the processed signal. Subsequently, the independent component analysis (ICA) based BSS algorithm is employed to separate the mixed signal preprocessed by the previous two ways. Finally, the proposed algorithm and the BSS detection method based on WT are compared in terms of spectrum analysis and separation performance. Simulation results show that the blind source separation detection method based on WT-SSA has a better signal detection performance.

DOI: 10.36244/ICJ.2025.4.6
Download 

 

T. Padmavathi, Kusma Kumari Cheepurupalli, and R. Madhu
Performance Evaluation of FBMC with optimal subcarrier spacing for 5G & beyond Communications
In 5G New Radio (NR) Release 15, the 3rd Generation Partnership Project (3GPP) Physical Layer Modulation for downlink and uplink communications, the Cyclic Prefix OFDM (CP-OFDM) is used. A wide range of potential use cases will describe future wireless networks. In order to achieve this, time-frequency resources must be dynamically assigned, which is difficult for traditional Orthogonal Frequency Division Multiplexing (OFDM). Therefore, OFDM improvements like filtering or windowing are needed. On the other hand, a multicarrier method like Filter Bank Multi-Carrier (FBMC) can be employed. Several prototype filters, including Hermite, PHYDYAS, and Root Raise Cosine (RRC), are used in this work to develop the framework for FBMC. Time-frequency efficiency will be determined for each user in the same band by adjusting the subcarrier spacing. The performance of the Signal to Interference Noise Ratio (SIR) is calculated for FBMC using varying subcarrier spacing and compared with different multicarrier transmission methods such as f-OFDM (filtered OFDM), CP-OFDM, UFMC (Universal Filter Multi Carrier), Weighted Overlap and Add (WOLA). FBMC outperforms CP OFDM, UFMC, f-OFDM, and WOLA in terms of Signal-to-Interference-plus-Noise Ratio (SIR), especially when subcarrier spacing is short (15 kHz, 30 kHz), where spectral leakage is most noticeable. The PHYDYAS filter performed better than the other ones, reducing intercarrier interference and increasing spectral efficiency by 20–30% even in asynchronous transmission scenarios. Furthermore, FBMC improved bandwidth economy by maintaining excellent performance without requiring a cyclic prefix. According to these indings, FBMC is a strong contender for upcoming 5G upgrades and 6G networks that require flexible waveform design, low outof- band emissions, and support for a variety of service classes, such as mMTC, URLLC, and non- orthogonal transmissions.

DOI: 10.36244/ICJ.2025.4.7
Download 

 

Sunil Kumar Shah, Raghvendra Sharma, and Neeraj Shukla
Dynamic XTEA Optimization and Secure Key Management for Embedded Microcontroller-Based SDN Systems in Smart Cities
The rapid expansion of smart city infrastructures necessitates robust and efficient security mechanisms for embedded processor-based Software-Defined Networking (SDN) nodes. Hence, this research introduces the Adaptive Secure  XTEA for Embedded Microcontrollers (ASX-EM), a novel encryption method designed to address these environments' unique security and performance needs. Existing encryption implementations often neglect proper padding validation, leading to vulnerabilities such as the Padding Oracle Attack (POA). The proposed Context-Aware Key Expansion and Secure Padding Validation (CAKE-SPV) technique customizes key scheduling based on node-specific parameters and employs a robust padding verification mechanism, significantly enhancing encryption security. Moreover, the computational demands of XTEA, with its multiple rounds of operations, are inefficient on resourceconstrained 8-bit microcontrollers, leading to increased latency and reduced system responsiveness. To optimize performance, the Adaptive Round and Parallel Processing (ARPP) method is developed that dynamically adjusts encryption rounds based on system metrics and employs bit-slice processing with precomputed lookup tables for efficient arithmetic operations. The results show that the proposed model has a low encryption time of 2.7s a decryption time of 2.6s, and a high encryption throughput of 22,000 KB/sec and, a decryption throughput of 17600 KB/sec, compared to other existing models.

DOI: 10.36244/ICJ.2025.4.8
Download 

 

Botond L. Márton, and László Bacsárdi
The Effect of Large Jumps in a Ring-like Quantum Network
Quantum communications promises major changes in today’s communication networks by sending qubits over long distances. These qubits enable large-scale quantum computing or information-theoretically secure distribution of symmetrical keys. One of the main enablers is quantum teleportation, which makes sending quantum information between two nodes possible even when they are far apart. From these nodes, one can build a larger quantum network, but due to the nature of quantum physics, certain tasks that are well understood in classical networks, such as routing, cannot be handled in a similar way. Our work focuses on modifying a previously created model for a ring-like quantum network and assessing the effect of introducing a new node type. Our results show that this node can alter properties of the underlying network. We also look at the possibilities of modeling the capacity of the network as well as the availability of the newly introduced edges, which open interesting questions for future research.

DOI: 10.36244/ICJ.2025.4.9
Download