In the beyond 5G/6G, the usage of the wideband transmission in the higher frequency band is expected. In this paper, we propose a method to compensate IQ imbalance caused by the path length differences between I and Q channels in the direct conversion transmitter in the wideband OFDM systems. In the proposed method, compensation signals are generated using the path length difference and the modulation symbols transmitted on the paired subcarriers which are known in the transmitter side. We analyzed the IQ imbalance and the proposed compensation method, then the effects of the proposed method are explained through the simulation.

Wireless communication systems require accurate channel estimation to compensate for the effects of fading channels. Traditional methods, such as minimum mean square error (MMSE) channel estimation, rely heavily on prior channel knowledge. Different types of MMSE channel estimation can be defined by the kinds of the prior channel knowledge applied. Recently, machine learning (ML) techniques, especially neural networks (NN), have shown great effectiveness in handling large volumes of data. This letter compares the performance of NN with various types of MMSE channel estimation. The findings confirm that NN can outperform MMSE methods that rely solely on statistical knowledge, maintaining high estimation accuracy even in severe frequency selective channel.

This paper describes an indoor location estimation method using the received signal strength indicators (RSSI) of wireless local area networks (LANs). Outdoor localization technology based on global navigation satellite systems (GNSSs) is currently used in many areas of our daily lives. However, the accuracy of GNSS is greatly degraded in underground malls and buildings. Such indoor location estimation includes methods that use Wi-Fi and Bluetooth RSSI and methods that use sensors in the terminal. For indoor localization, the present study uses the fingerprint method, which estimates RSSI emitted from wireless LAN access points (APs) by adapting a convolutional neural network (CNN). However, the fingerprinting method has a drawback, which is that the RSSI is generally attenuated by the body’s shadow depending on the direction it is measured from. For this problem, by utilizing the directional information at the time of measurement for both learning and estimating, it is possible to make estimations taking into consideration the directional context. In this study, we proposed fingerprinting localization using a multi-input CNN model and directional information. We reported in ICETC2024 that accuracy improved as the number of input azimuths increased. However, during the verification process, there were cases where the number of input azimuths did not correspond to the improvement in accuracy. Therefore, we conducted further investigation into this issue. Another point to consider in the multi-input model is the effect of changes in the input layers on accuracy. To examine this, we conducted verification by changing the input layer of the directional information and confirmed that accuracy improved.

Higher frequency bands are essential for next-generation wireless local area network (LAN). Due to the significant propagation loss of sub-terahertz band, beamforming is indispensable. However, since beamforming is affected by obstructions, stable communication requires simultaneous transmission with multiple access points (APs). In a multi-AP environment, although optimal resource scheduling can be determined through exhaustive search, the complexity is quite high. In this letter, we evaluate resource scheduling using self-supervised learning, which does not rely on teacher labels for training. By computer simulation, we show that self-supervised learning can achieve comparable to that of exhaustive search.

Recently, TV broadcasters worldwide have begun offering internet linear streaming services in addition to traditional broadcasting services. This study proposes a cloud-based architecture for integrating television broadcasting and internet linear streaming. This architecture utilizes cloud-computing resources and an optical path network assuming an All-Photonics Network (APN) on demand, according to the program schedules. In addition, the streaming system on this architecture provides quick emergency program insertion and personalized services by utilizing edge computing on a content delivery network (CDN). This paper also describes the verification results of a prototype system implemented on the public cloud and the CDN.

With the increasing demand for network services, energy efficiency in Passive Optical Networks (PONs) has become a critical concern. This study addresses how to decrease power consumption of the Optical Network Unit (ONU) by optimizing its laser diode driver (LDD) circuit. A CMOS-based distributed amplifier (DA) is proposed, incorporating low output matching impedance, choke inductors, and gate line-division techniques. We numerically demonstrate a power consumption of 300.7mW, representing a 68% reduction compared to conventional designs, while maintaining a 43-mA peak-to-peak differential current output and a -3-dB bandwidth of 2.2GHz. These findings demonstrate the potential of CMOS-based DAs for developing energy-efficient LDDs in modern access networks.

In the event of a large-scale disaster, effective communication is crucial to verify the safety of those affected. However, disruption of the communication infrastructure makes it difficult to collect information on affected areas. Therefore, a rapid restoration of the communication network is required. In this letter, we focus on the realization of a low transmission delay drone-based wireless relay network as an emergency communication system. Our study investigates how wind direction and speed affect the information transmission time of a conventional drone-based network. Subsequently, we introduce an innovative flight model designed to reduce wind effects and assess its delay performance.

Belief propagation decoding of Polar codes suffers from performance degradation if the frozen bits are not appropriately allocated. Particularly, when simplified log-likelihood ratio (LLR) calculations are used to reduce hardware complexity, the conventional frozen bit allocation based on protograph extrinsic information transfer (PEXIT) analysis becomes ineffective due to the inaccurate EXIT function. In this paper, we propose a new EXIT function that takes into account the approximation in the LLR for PEXIT analysis. Simulation results show that PEXIT analysis with the proposed EXIT function is able to represent the degradation of bit error rate (BER) due to the LLR approximation.