論文

This paper investigated the impact of a quantizer in analog-to-digital and digital-to-analog converters in communication devices on image quality when using deep learning-based joint source-channel coding modulation (JSCCM) for image transmission. In recent years, JSCCM, which efficiently encodes images and videos with low information entropy, has attracted great attention. JSCCM has a structure based on an autoencoder and determines the compression ratios for the image input by adjusting the number of IQ symbol output. The IQ symbol output from the encoder are allocated to symbol constellations with higher degrees of arbitrariness than those in typical square quadrature amplitude modulation and are therefore expected to be strongly affected by the quantization noise. In this paper, we employed quantization to the IQ symbol sequence and investigated its effect. Adjusting the quantizer's clipping ratio and the number of quantization bits, we examined the images' tolerance of the peak signal-to-noise ratio (PSNR). The simulation results showed that by adequately adjusting the clipping ratio, the image quality can be guaranteed to be equivalent to ideal conditions without quantization noise, and the number of required quantization bits that do not degrade the PSNR, was calculated.

This paper investigates the impact of time-varying channels on deep joint) source-channel coding (Deep JSCC over multiple-input multiple-output (MIMO) systems. While Deep JSCC is known for its robustness to noise, its resilience to time variations in mobile environments remains underexplored. In this paper, we first train the autoencoder using a single-input single-output (SISO) transmission under an AWGN channel on the premise that MIMO channel is equalized separately. During testing, images are transmitted over time-varying MIMO channels with various precoding techniques. The received signals are processed through a MIMO precoder before decoding. We evaluate image quality using PSNR and SSIM under various MIMO configurations. The results reveal that time variations significantly degrade Deep JSCC performance, particularly at higher mobility speeds. Its coutermeasure is increasing the number of receiving antennas that can effectively mitigates impact of channel aging.

Deep joint source-channel coding (JSCC) has been attracting attention for achieving task-oriented communication. It replaces traditional information source coding and channel coding with a deep learning-based autoencoder, directly mapping information sources such as images to IQ symbols. For images, it is claimed to avoid the cliff effect and achieve a higher peak signal noise ratio (PSNR) even in low SNR regions. While related work has assumed various propagation channel models and validated the effectiveness of Deep JSCC, there are few reports confirming its principles through experiments. Specifically, to the best of our knowledge, there are no reported examples of experiments of Deep JSCC in 5G systems. In this paper, we present a proof-of-concept of Deep JSCC in a 5G system. We modified commercially available 5G base stations (gNB) and 5G terminals to enable input and output of IQ data from external devices. We connect the 5G devices using coaxial cables and attenuators, transmit and receive JSCC signals, and evaluate the PSNR. The results demonstrate that even when communicating at power levels lower than the minimum receiver sensitivity specified in the receiver's datasheet, the image can be successfully restored with less than 1 dB degradation in PSNR compared with the simulation result.

Introduction: Eigenvalue-modulated signals have invariant carriers even after propagating through the nonlinear fiber channel. The eigenvalues are extracted based on inverse scattering transform (IST) [1]. IST makes a matrix and it is solved by using the QZ method. Then, the QZ outputs the eigenvalues in order of maximum absolute value in the almost case. In the demodulation of the on-off encoded eigenvalue signal, we have proposed employing neural network (NN) on the bit decision [2]. The NN is sensitive to the input order while the higher optical signal-to-noise ratio (OSNR) gain is obtained than the common hard decision. This paper proposes employing deep sets layer on neural eigenvalue demodulation. This paper reports the experimental result of the 50-km fiber transmission.

This paper proposes a novel blind adaptive array (BAA) interference suppression scheme incorporated with a time-domain symbol spreading (TISS). It can expand an operational region of BAA irrespective of input signal-to-interference power ratio (SIR) conditions. Independent component analysis (ICA) and constant modulus algorithm (CMA) are expected as the promising BAA approaches to suppress unknown interference which is extensively caused due to the small cell densification and spectrum superposing between a plurality of wireless communication systems. These BAA algorithms basically require an input SIR to be larger than 0 dB to suppress interference by capturing the desired signal appropriately. The proposed scheme applies TISS to the transmission signal for the desired user and then de-spreads it at the receiver side. It Gaussianizes statistical inherence of the interference signals and can bring beneficial effects; reducing the kurtosis for ICA as well as collapsing the constant envelope properties for CMA. Such intentional modification to the interference signals can improve the capture performance of the desired signal and realize its effective identification even at SIR < 0 dB. Its fundamental effectiveness is presented through various perspectives such as input SIR, symbol amount, and modulation orders.

Blind adaptive array (BAA) can suppress unknown co-channel interference without reference signal. Vehicle-to-everything (V2X) communication and underwater acoustic communication (UWAC) are seen as promising application towards Beyond 5G and 6G. Capacity enhancement is still essential for such extensive scenario and thus effective interference suppression can contribute to realizing efficient spatially multiplexed transmission without frequent reference signal insertion. Focusing on the constant modulus algorithm (CMA) as one of the BAAs, this paper investigates its fundamental interference suppression performance under the intense time varying channel environment, anticipating UWAC application. In addition, effectiveness of time-domain inter-symbol spreading (TISS) to expand operating domain of CMA is also clarified under several interference environment.

This paper proposes the suppression scheme of interference from backward reflection on single-wavelength bidirectional transmission in optical access network systems using asymmetric symbol rate. We obtained 2.6-dB gain of signal-to-interference power ratio at 2.5-fold symbol rate ratio for the QPSK signal.

We demonstrate a joint demodulation scheme using a parallel eigenvalue solver and ANN. The proposed scheme is demonstrated experimentally by 2000-km fiber transmission with < 2-dB OSNR penalty compared with conventional QZ method.

We demonstrate a joint demodulation scheme using a parallel eigenvalue solver and ANN. The proposed scheme is demonstrated experimentally by 2000-km fiber transmission with < 2-dB OSNR penalty compared with conventional QZ method.

Optical camera communication (OCC) can be realized by commercial LEDs or displays as a transmitter and image sensors as a receiver. One of the challenges to enhance the transmission capacity in OCC is a two-dimensional light source with a display at the transmitting side. So far, the optimization of the imaging process and the method of tracking and detecting the light source have not been studied in detail. This paper proposes a dynamic light source detection system based on the A-QL method specified in IEEE 802.15.7 as a transmission symbol format. It employs YOLO, a deep learning-based object detection algorithm, and optimizes the image capture process as a symbol detection system. The proposed system can detect two-dimensional symbols with adjusting its angle and orientation. Its effectiveness and feasibility are demonstrated through an experimental evaluation.

The Internet of Things (IoT) has been growing rapidly in recent years. With the appearance of 5G, it is expected to become even more indispensable to people's lives. In accordance with the increase of Distributed Denial-of-Service (DDoS) attacks from IoT devices, DDoS defense has become a hot research topic. DDoS detection mechanisms executed on routers and SDN environments have been intensely studied. However, these methods have the disadvantage of requiring the cost and performance of the devices. In addition, there is no existing DDoS mitigation algorithm on the network edge that can be performed with the low-cost and low-performance equipment. Therefore, this paper proposes a light-weight DDoS mitigation scheme at the network edge using limited resources of inexpensive devices such as home gateways. The goal of the proposed scheme is to detect and mitigate flooding attacks. It utilizes unused queue resources to detect malicious flows by random shuffling of queue allocation and discard the packets of the detected flows. The performance of the proposed scheme was confirmed via theoretical analysis and computer simulation. The simulation results match the theoretical results and the proposed algorithm can efficiently detect malicious flows using limited resources.

Optical eigenvalue transmission based on inverse scattering transform (IST) has been studied for one of approaches to overcome the Kerr nonlinearity limit in fiber optic communications. In the recent decade, several multilevel modulation schemes that are based on IST, such as 16-ary and 64-ary signals using on-off encoding and b-modulation of multieigenvalues, have been proposed. To increase the transmission capacity and extend the transmission distance, applications of machine learning-based approaches to IST-based transmission have been proposed and demonstrated. Another prospective approach to increase the transmission capacity and extend the transmission distance in fiber optic communication involves the application of soft-decision forward error correction (SD-FEC). However, the applicability of SD-FEC to eigenvalue-modulated signals has not yet been investigated in detail because the distribution of the received signal is complicated for IST-based transmissions. In this paper, we describe in detail the theory of optical eigenvalue transmission, including the design of multilevel eigenvalue-modulated signals and the effects of noise and scaring parameters (coefficients for normalization of the nonlinear Schrödinger equation). We explain why neural network (NN) demodulators are advantageous for eigenvalue transmission systems. Consequently, we propose a combination of NN-based demodulators and SD-FEC decoding. A multilabel NN-based demodulator is employed to compute the L-value from the received eigenvalue pattern at the receiver. For a 16-ary eigenvalue-modulated signal, the proposed method outperformed a combination of the Gaussian approximation and SD-FEC in the simulation. Moreover, the experimental results show successful operation with error-free transmission through a 3000-km optical fiber line. In addition, we experimentally demonstrate the applicability of SD-FEC to a 4096-ary eigenvalue-modulated signal. The experimental results indicate that an achievable transmission distance can be extended to 1200 km using the NN demodulator and SD-FEC.