13th Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)

DECEMBER 14 – 17, 2021, TOKYO, JAPAN
Venue: KFC Hall & Rooms
Kokusai Fashion Centre Bldg., Yokoami 1-6-1, Sumida City, Tokyo

Signal & Information Processing — Science for Signals, Data, and Intelligence

Important Dates

[April 1, 2021] Submission of Proposals for Special Sessions [May 1, 2021] Submission of Proposals for Forum, Panel & Tutorial Sessions [July 15, 2021] Submission of Regular Papers [July 15, 2021] Submission of Special Session Papers [July 16 to September 10, 2021] Submission of Research Abstract  [August 31, 2021] Notification of Papers Acceptance [October 1, 2021] Submission of Camera‑Ready Papers [October 1, 2021] Author (Early-Bird) Registration Deadline [December 14 – 17, 2021] Tutorials, Summit and Conference Dates

Special Session

Session # Target track Session title
SS-01SPSHigh-Performance Image Processing
SS-02Reconfigurable Computing and Performance Evaluation
SS-03Complex- and Hypercomplex-valued Adaptive Signal Processing
SS-04Digital Filters and Its Applications
SS-05Deep Learning Systems and Applications for Cloud, Fog, and Edge Computing
SS-06SIPTMSignal Processing and Machine Learning over Graphs
SS-07Recent Advances in Acoustic & Biomedical Signal Processing
SS-08Objective Measurements in Psychological and Cognitive Sciences
SS-09SLARobust Speaker Recognition with Microphone Arrays
SS-10Multimodal Learning for Biological and Biomedical Acoustic Signal Processing
SS-11Advanced Signal Processing and Machine Learning for Audio and Speech Applications
SS-12Recent Topics on Signal and Information Processing for Active Control of Sound
SS-13Advanced Topics on Sound Event and Scene Analysis
SS-14 BioSiPSDeep Learning for Biomedical Signal Processing and Systems
SS-15Signal Processing in Behavior Analysis
SS-16 IVMData Hiding Techniques for Enriched Multimedia and Beyond
SS-17Advances in Human Behavior Sensing and Understanding
SS-18High Performance Intelligent Technologies for Image and Video Applications
SS-19Advanced Image and Video Processing using Deep Learning
SS-20Recent Advances in Deep Image Restoration
SS-21Advanced Topics in Low Precision Image Processing
SS-22High Performance Image and Video Processing and Applications
SS-23MSFDeep Generative Models for Media Clones and Its Detection
SS-24Multimedia Security and Deepface
SS-25Intelligent Approaches in Signal Processing for Multimedia Security
SS-26The Future of Biometrics beyond Recognition and Security
SS-27WCNDigital Convergence of 5G/B5G, AIoT and Security
SS-28Advanced Wireless Access Technologies and Data Analysis for IoT and Environmental Monitor
SS-29MLDADeep Learning for Few-shot Data Analysis
SS-30Online and Distributed Kernel Learning Algorithms

Target track

  • - Signal Processing Systems: Design and Implementation (SPS)
  • - Signal and Information Processing Theory and Methods(SIPTM)
  • - Speech, Language, and Audio (SLA)
  • - Biomedical Signal Processing and Systems (BioSiPS)
  • - Image, Video, and Multimedia (IVM)
  • - Multimedia Security and Forensics (MSF)
  • - Wireless Communications and Networking (WCN)
  • - Machine Learning and Data Analytics (MLDA)

*Replace “at” in the e-mail address below with @.

SS-01

Target track
SPS
Title
High-Performance Image Processing
Organizer(s)
Norishige Fukushima
Contact info.
fukushima at nitech.ac.jp
Abstract
This session includes high-performance computing for image processing. For example, the followings are main topics for this session, but not are limited.
  • - Approximated computing for fast image processing
  • - Approximated algorithm for fast image processing
  • - Programming language for image processing
  • - Domain specific language for image processing
  • - Compiler for image processing
  • - Code optimization for image processing
  • - Parallel computing for image processing
  • - Real-time implementation of image processing application
  • - Hardware implementation of image processing application

SS-02

Target track
SPS
Title
Reconfigurable Computing and Performance Evaluation
Organizer(s)
Junyong Deng, Xiaoyan Xie
Contact info.
djy at xupt.edu.cn
Abstract
There are many types of algorithms for multimedia applications with different applicable scenarios, and large differences in performance. The reconfigurable architecture adopts a data parallel structure and has the advantages of low power consumption, high performance, safety, flexibility, parallelism, and low cost. It has become a promising platform for most multimedia applications. This session focused on but not limited to the following areas:
  1. 1. Reconfigurable architecture for certain application areas
  2. 2. Reconfigurable mechanism
  3. 3. High-performance computing methods
  4. 4. Performance characterization and image recognition optimization
  5. 5. Algorithm optimization of video encoding
  6. 6. Algorithm, implementation and application of deep learning

SS-03

Target track
SPS
Title
Complex-and Hypercomplex-valued Adaptive Signal Processing
Organizer(s)
Akira Hirose, Isao Yamada
Contact info.
ahirose at ee.t.u-tokyo.ac.jp
Abstract
Complex-and hypercomplex-valued neural networks atract a variety of researchers in sensing and imaging fields. This session aims to discuss latest progress in the fields in the theory and applications.

SS-04

Target track
SPS
Title
Digital Filters and Its Applications
Organizer(s)
Shunsuke Yamaki, Kenji Suyama
Contact info.
yamaki at tohoku.ac.jp
Abstract
Digital filter technology is an important topic in the signal processing field despite a large amounts of works already have been reported. This session consists of the recent research topics on digital filter design and digital filter applications.

SS-05

Target track
SPS
Title
Deep Learning Systems and Applications for Cloud, Fog, and Edge Computing
Organizer(s)
Jia-Ching Wang, Pao-Chi Chang
Contact info.
jcw at csie.ncu.edu.tw
Abstract
Deep learning has achieved great success in numerous perceptual tasks (e.g., object detection, image understanding, and speech recognition). At the same time, the number of end devices, such as IoT (Internet of Things) devices, has dramatically increased. These devices usually aim at some machine learning-based perceptual tasks or applications as they are often directly connected to sensors (e.g., cameras) that continuously capture a large quantity of input data. However, traditional cloud-based infrastructures have been not enough for the demands of the current state of machine learning systems on end devices due to some limitations. Some major issues include the associated communication costs, latency, security, and privacy concerns induced by current cloud-based infrastructures. Therefore, the concepts of fog and edge computing have been recently proposed to alleviate these limitations by moving data processing capabilities closer to the network edge. This special session will focus on all aspects of deep learning network architectures, algorithms, and applications, with particularly emphasizing on exploring recent advances in deep neural networks over the cloud, fog, edge, and end devices. Topics of interest include, but are not limited to: . Distributed Deep Neural Networks . Cloud, Fog, and Edge Computing with Deep Learning Systems . Security and Privacy over Cloud, Fog, Edge, and End Devices with Deep Learning Systems . Sensor Fusion on Different End Devices with Deep Neural Networks . Jointly Model Learning over Cloud, Fog, Edge, and End Devices with Deep Learning Systems . General Deep Learning Applications in Any Aspects of Perceptual Tasks (e.g., Object/Face Detection and Recognition, Image/Video Understanding, and Audio/Speech Recognition)

SS-06

Target track
SIPTM
Title
Signal Processing and Machine Learning over Graphs
Organizer(s)
Yuichi Tanaka, Shunsuke Ono
Contact info.
ytnk at cc.tuat.ac.jp
Abstract
With the current advancement of sensing technologies, signals and features often distribute irregularly in space for a wide range of applications of signal processing, sensor networks, data mining, and machine learning, to name a few. Researchers have therefore focused on investigating theory and algorithms for signal processing and machine learning on networks. Networks are mathematically represented by graphs: Graph signal processing and graph neural networks have been a hot topic for a decade while we still have various open problems both in theory and applications. In this special session, we invite active researchers in this field to discuss the high-level ongoing studies on signal processing/machine learning on graphs, from a fundamental theory to practical applications. We believe that this special session is timely, and fruitful discussions among these experts and audience may lead to many new ideas and collaborations.

SS-07

Target track
SIPTM
Title
Recent Advances in Acoustic & Biomedical Signal Processing
Organizer(s)
Kiyoshi Nishikawa, Muhammad Tahir Akhtar, Prashant Kumar Jamwal
Contact info.
kiyoshi at tmu.ac.jp
Abstract
The main focus of the proposed session is to invite contributions in the area of Acoustic and Biomedical Signal Processing. The proposals are invited for signal processing and machine application in the biomedical area, with focus on both theoretical as well as practical research advances. The target applications include, but not limited to, acoustic signal processing, artificial intelligence based algorithms for biomedical image processing and data analytics, linear and nonlinear adaptive signal processing, applications-led research on the use of methods and devices in clinical diagnosis, patient monitoring and management. etc. Prospective authors are encouraged to consider submitting a manuscript to this special session.

SS-08

Target track
SIPTM
Title
Objective Measurements in Psychological and Cognitive Sciences
Organizer(s)
Toshihiko Matsuka, Yoshiko Kawabata
Contact info.
matsuka.toshihiko at gmail.com
Abstract
Researches in psychological and cognitive sciences often use subjective measures of, for example, beliefs and evaluations of targeted objects, entities, events, for so forth. Although this method is perfectly valid, great care is needed to attain reasonable levels of reliability. On the other hand, there are several methods in these fields using objective measures such as brain and other physiological activities. The present session focus on recent applications or developments of objective measurements in research areas that traditionally uses subjective measurements. In particular, this proposed special session invites research on methods that offer researchers to conduct behavioral experiments with objective measures of humans' beliefs, attitudes, and/or evaluations. The proposed special session focuses on recent advances in theories and methods that allow us to conduct up-to-date novel research on human psychological and/or cognitive processes. In addition, it also aims specific applications of those theories and methods to behavioral experiments. The areas involved including learning, perception, natural language processing, decision-making.

SS-09

Target track
SLA
Title
Robust Speaker Recognition with Microphone Arrays
Organizer(s)
Xiao-Lei Zhan, Qingyang Hong
Contact info.
xiaolei.zhang at nwpu.edu.cn
Abstract
Speaker recognition is a task of identifying persons from their voices. Recently, deep learning has dramatically revolutionized speaker recognition. Along with the rapid progress of speaker recognition, the frontier turns to "Rerecognition in the wild", where lots of channel/speaker mismatch and noisy problems arise. To address the issue, recently, there is a tremendous need and a clear research trend for the improvement of speaker recognition with microphone arrays, given strong application demand in internet of things, smart home, and smart city. To overcome these difficulties, many domain adaptation and noise reduction methods based on microphone arrays were proposed. However, the performance is far from satisfactory. For example, the 5th and 6th CHiME Speech Separation and Recognition Challenge (CHiME-5 and CHiME-6) has recognized that the performance of speaker recognition is the main problem that accounts for the high speech recognition error rates of the baseline systems. To address the above problem, it is strongly needed to organize a special session on speaker recognition with microphone arrays, which gathers researchers from both speaker recognition and speech enhancement for a deep technique discussion on new ideas and directions.

SS-10

Target track
SLA
Title
Multimodal Learning for Biological and Biomedical Acoustic Signal Processing
Organizer(s)
Yu Tsao, Shih-Hau Fang, Ying-Ren Chien, Syu-Siang Wang, Kai-Chun Liu
Contact info.
yu.tsao at citi.sinica.edu.tw
Abstract
Due to the recent advance of sensing, computing, and communication capabilities, vast amounts of multimodal data can be readily accessed. With such large amount and a wide diversity of multimodal data, systems for various applications can be constructed based on the state-of-the-art machine learning algorithms. Previous works have revealed that integrating heterogenous data from multimodalities can facilitate improved classification and regression performance for specific target tasks. Meanwhile, numerous multimodality systems with novel data and system architectures have also been developed. The goal of this special session is to further advocate this research direction of using multimodal data to leverage the prediction accuracies, information retrieval, and enhancement/separation for biological and biomedical acoustic signals. We look forward novel systems that integrate multiple modalities, novel architectures, and feature processing. We also encourage submissions that discuss practical issues of multimodality data recording/capturing and system designs along with potential solutions through this special session.

SS-11

Target track
SLA
Title
Advanced Signal Processing and Machine Learning for Audio and Speech Applications
Organizer(s)
Shoji Makino, Hiroshi Saruwatari
Contact info.
s.makino at waseda.jp
Abstract
We are surrounded by sounds in our daily lives. To achieve a rich acoustic environment for the next generation, audio and acoustic signal processing technologies are essential. The audio and acoustic signal processing include (but are not limited to) source localization, source separation, dereverberation, noise reduction, and virtual acoustic reproduction. These techniques form the core of the state-of-the-art audio and acoustic signal processing and are indispensable to the realization of future communication via for both man-machine and human-human interfaces. This special session is dedicated to advanced signal processing and machine learning for audio and speech applications. The aim of this special session is to offer an opportunity to link these techniques in different areas and to find effective ways of achieving our goals. This special session represents a vehicle whereby researchers can present new studies, thus paving the way for future developments in the field. This special session will stimulate interest in the challenging area of audio and acoustic signal processing, and create an increasing body of high-quality research aligned with this idea.

SS-12

Target track
SLA
Title
Recent Topics on Signal and Information Processing for Active Control of Sound
Organizer(s)
Yoshinobu Kajikawa, Chuang Shi
Contact info.
kaji at kansai-u.ac.jp
Abstract
We are seeing an increase in performance and new features in today's audio and acoustics systems in consumer applications that are enabled by the latest signal processing algorithms. With the ever-demanding consumer market, signal processing will continue to excel in many applications, such as noise cancellation, signal enhancement, sound effect generation, audio content searching, smart sound control, sound profiling and equalization, psychoacoustic enhancement, personal audio, and many others. This special session aims to put together papers that focus on the new advances on signal and information processing for active control of sound, with emphasis on successful deployment and application of these new algorithms. Active control of sound includes active noise control (ANC), sound image localization, spatial audio systems, and so forth.

SS-13

Target track
SLA
Title
Advanced Topics on Sound Event and Scene Analysis
Organizer(s)
Nobutaka Ono, Keisuke Imoto, Tatsuya Komatsu
Contact info.
onono at tmu.ac.jp
Abstract
We are surrounded by various kinds of sounds such as speech, music, and environmental sounds. To understand these sounds by computers and realize human-like listening systems, sound event and scene analysis are essential technologies, and they have been extensively developed. This special session is dedicated to recent advances in sound event and scene analysis, feature extraction methods for sound event/scene analysis, event/scene modeling methods based on deep learning, and applications such as monitoring elderly people, surveillance, life-logging, and advanced multimedia retrieval.

SS-14

Target track
BioSiPS
Title
Deep Learning for Biomedical Signal Processing and Systems
Organizer(s)
Yibin Tang, Aimin Jiang, Hon Keung Kwan
Contact info.
tangyb at hhuc.edu.cn
Abstract
Artificial intelligence has been successfully applied in various medical scenarios such as biomedical signal analysis and medical system design. Particularly, deep learning methods are an effective tool to tackle the complex problems. It learns the intrinsic attributes beyond the given biosignals or other related physical sensor observations. However, practical problems remain unsolved on both methodology and engineering applications, which seriously impacts the performance of accurate and intelligent diagnosis. The proposed session is devoted to the publication of high-quality papers for accurate diagnosis and rehabilitation system. Deep learning and signal processing methods are encouraged in this session to achieve impressive results.

SS-15

Target track
BioSiPS
Title
Signal Processing in Behavior Analysis
Organizer(s)
Kazushi Ikeda, Toshitaka Yamakawa
Contact info.
kazushi at is.naist.jp
Abstract
Thanks to the recent IoT technology, behavioral data such as life-log of not only human but also animals have got easier to collect. Such massive data include valuable information and new signal processing techniques such as machine learning have been developed. This special session is planned to gather recent techniques for behavioral data analysis and to accelerate their improvement.

SS-16

Target track
IVM
Title
Data Hiding Techniques for Enriched Multimedia and Beyond
Organizer(s)
Tetsuya Kojima, Masashi Unoki
Contact info.
kojt at tokyo-ct.ac.jp
Abstract
Data hiding technology has been frequently utilized as digital watermarking or digital finger printing to protect intellectual property rights as well as to avoid illegal copying or distribution of multimedia contents. An alternative objective of data hiding is to give additional values to multimedia works. This can make the multimedia works more attractive and versatile. There have been lots of studies from this point of view by utilizing image, audio and video data hiding techniques. In this special session, we explore the novel aspects of data hiding techniques for multimedia contents and digital entertainments in our future society.

SS-17

Target track
IVM
Title
Advances in Human Behavior Sensing and Understanding
Organizer(s)
Jianquan Liu, Albert Ali Salah, Mohan S. Kankanhalli
Contact info.
jqliu at nec.com
Abstract
With the development of AI technology and IoT devices, research and development related to information processing technology that handles images, videos, sounds, sensors, etc. is becoming more active than before. In various application situations such as urban safety and smart city, the development of technology from different perspectives is required to analyze human behavior. For example, how to recognize human behavior at high speed and high accuracy in multi-sources such as images, videos, and sounds, how to analyze and find human behavior from a large amount of data at high speed and high accuracy, and how to summarize the actions of people on the axis of time and space, and automatically tell the story of who, when, where, and what they did. In addition, since it is a processing technology related to "humans," it is important to consider how to provide the technology with a privacy-friendly mechanism. This special session will provide an academic forum for active discussions focusing on these contents, and will contribute to the future development of academia and industry.

SS-18

Target track
IVM
Title
High Performance Intelligent Technologies for Image and Video Applications
Organizer(s)
Jing-Ming Guo
Contact info.
jmguo at mail.ntust.edu.tw
Abstract
Recent advancements in Image and Video processing methods resulted in extensive surge of interest to the evolving artificial intelligence technology. This special session provides an excellent opportunity for research experts to present their innovative works in high performance video and multimedia applications. The critical objective is to develop unprecedented high accuracy image/video processing systems which are of significant engineering importance. The session emphasis on extensive imaging solutions meant for segmentation, classification, reconstruction, denoising, compression, security, human interaction, halftoning and biomedical application. Although many notable works are proposed, the deep neural networks and computer vision algorithms are not thoroughly exploited to its fullest efficacy. Specifically, there is a promising scope of improvements in-terms of its feasibility, adaptability and generalization towards many real-time applications. In summary, the special session covers advanced general topics related to challenging problems in image and video processing
  • - Instance and Semantic Segmentation
  • - Image Classification and Reconstruction
  • - Image and Video Compression
  • - Multimedia Security and Watermarking
  • - Human Interactions and Smart Devices
  • - Imaging and Computer Graphics
  • - Biomedical Image Processing
  • - Deep innovative models (CNNs, DNNs, RNNs, GANs and Capsule Networks)
  • - Image Denoising and Quality Assessment

SS-19

Target track
IVM
Title
Advanced Image and Video Processing using Deep Learning
Organizer(s)
Je-Won Kang, Jae-Young Sim
Contact info.
jewonk at ewha.ac.kr
Abstract
This special session covers advanced image and video processing techniques based on deep learning. The topics include image compression, image enhancement, pose estimation, anomaly detection, face interpolation, and person search. In these topics, advances are being made at a fast pace, especially due to deep learning and availability of big data. By presenting these techniques, the session intends to provide the place for discussing new trends in image and video processing.

SS-20

Target track
IVM
Title
Recent Advances in Deep Image Restoration
Organizer(s)
Jong-Ok Kim
Contact info.
jokim at korea.ac.kr
Abstract
In recent years, deep learning techniques have been popularly adopted for traditional image restoration tasks, and the performances have been significantly enhanced due to their intelligent learning capability. In this special session, we invite four technical papers which are in the fields of image deblurring, denoising, and super-resolution. We are going to discuss benefit, limitation, and future direction of deep learning based image restoration techniques.

SS-21

Target track
IVM
Title
Advanced Topics in Low Precision Image Processing
Organizer(s)
Gwanggil Jeon, Shogo Muramatsu
Contact info.
gjeon at inu.ac.kr
Abstract
Cloud computing is a well-used approach for running diverse data-analysis methods to obtain insight from massive raw data. However, the current cloud computing system suffers from large computation/communication overhead while processing big data. Recently, edge computing has become an important way to handle the ever-increasing volume of user-generated data. The meaning of edge computing is to process data close to where it is collected rather than on the cloud. The edge computing paradigm in image processing comes with many challenges in terms of reducing complexity and processing running time while maintaining accuracy and specificity. In this special section, we will explore the issues of image, video, and multimedia processing at the edge and seek contributions addressing image, video, and multimedia topics.

SS-22

Target track
IVM
Title
High Performance Image and Video Processing and Applications
Organizer(s)
Kosin Chamnongthai
Contact info.
kosin.cha at kmutt.ac.th
Abstract
Image and video processing plays important role as basic process for information and knowledge processing. Researches and studies in the level of information and knowledge basically depend upon many tools and modules based on image and video processing. Although researches in information and knowledge processing have been undertaken actively, image and video processing should be improved and developed in term of performance in order to enhance quality of information and knowledge processing systems. This special session collects some cases of high performance image and video processing and their applications in pixel and frame levels.

SS-23

Target track
MSF
Title
Deep Generative Models for Media Clones and Its Detection
Organizer(s)
Ngoc-Dung T. Tieu, Minoru Kuribayashi
Contact info.
dungtieu at nii.ac.jp
Abstract
Fake images/videos, fake voices, fake news/reviews that are generated by machines have been proved incredibly successful at fooling both humans and authentication systems. The rapid growth in deep generation techniques for producing such fake media information may cause risks in our modern lives. Automatic detection of fake media information thus has become an essential issue. In addition to this, real media information anonymization is also an important topic to protect this information from being maliciously exploited to generate fake media information. Therefore, we promote a challenging topic for the development of generation and detection methods for fake media information.

SS-24

Target track
MSF
Title
Multimedia Security and Deepface
Organizer(s)
Xiangui Kang, Hitoshi Kiya, Jiantao Zhou, Ren Yanzhen
Contact info.
isskxg at mail.sysu.edu.cn
Abstract
While many traditional approaches are investigated to provide information security, machine learning, as the driving force of this wave of AI, provides powerful solutions to many real-world technical and scientific challenges. Multimedia security-oriented applications has recently received a new boost thanks to the development of powerful methods relying on the advancements of in particular deep learning (DL). On the other hand, AI has raised the problem of the counterfeiting of multimedia data to an unprecedented level. High quality fake videos and audios generated by AI algorithms (the deep fakes) have started to challenge the status of videos and audios as definitive evidence of events. Together with the problems raised by the inherent security of deep networks, new serious threats of vulnerability and fragility are then posed, with the consequent need for advanced systems capable of working under more and more challenging conditions. This special session aims at drawing the attention of researchers towards the new challenges posed by the use of AI in multimedia security, including Deepfake generation and its detection, the susceptibility to adversarial attacks, the need for a huge amount of labeled training data, the risk of data overfitting with consequent failures in the presence of unforeseen situations at test time, etc. This special session aims to put together papers that focus on Deepfake generation and its detection, exploiting DL methods for a wide variety of multimedia security applications, investigating the more general problem of a systematic development of secure AI tools, the novel approaches capable of overcoming the limitations of state-of-the-art DL methods while keeping the superiority of modern AI tools.

SS-25

Target track
MSF
Title
Intelligent Approaches in Signal Processing for Multimedia Security
Organizer(s)
SimYing Ong, Tew Yiqi, KokSheik Wong
Contact info.
simying.ong at um.edu.my
Abstract
This session contributes new algorithms to ensure secured communications and prevent unauthorized information exchange in secured multimedia systems. It focuses on numerous applications' algorithms and scenarios. In addition, it offers an in-depth analysis of information hiding techniques including cryptography, watermarking, encryption/decryption, data embedding and authentication. As part of the session's comprehensive coverage, potential discussion on contemporary multimedia authentication and fingerprinting techniques, fine-tuning on watermarking model, improved edge estimation in reversible data hiding method and more interesting multimedia security topics to be presented. This session also equips perception technology professionals with the latest technologies, approaches, and techniques for multimedia security systems, offering a valuable resource for researchers working to develop secured signal processing systems.

SS-26

Target track
MSF
Title
The Future of Biometrics beyond Recognition and Security
Organizer(s)
Koichi Ito, Tetsushi Ohki
Contact info.
ito at aoki.ecei.tohoku.ac.jp
Abstract
Biometrics uses biological or behavioral characteristics to authenticate a person and has attracted much attention as a new authentication approach against traditional ones such as key, password, etc. Recent developments in biometric technology have achieved high authentication accuracy, security, and convenience, but despite these advances, there still remains many challenges, which we can chart in many directions. This special session will focus on these issues and directions related to recognition and security. It covers a wide range of areas including face recognition and other authentication modalities, security such as template protection, anti-spoofing etc.

SS-27

Target track
WCN
Title
Digital Convergence of 5G/B5G, AIoT and Security
Organizer(s)
Wen-Ping Lai, Po-Chiang Lin, Chung-Nan Lee
Contact info.
wpl at saturn.yzu.edu.tw
Abstract
Differential service needs have been the driving force for 5G/B5G to serve as an innovative platform for digital convergence of information, control and management, based on cutting-edge technologies such as software defined networking, network functions virtualization and multi-access edge computing. By connecting everyone to everything, the 5G/B5G technology is expected to bring about a new industrial and technological revolution. With the various challenges caused by the enormous number of devices connected through 5G/B5G, AIoT (AI + IoT) is a revolutionary combination that would provide promising solutions, such as low-latency from-core-to-edge intelligence for connected vehicles/drones and tactile internet. As billions of devices use the 5G/B5G radio access network (RAN), it will increase the risk of RAN resources overloaded by some attacks, such as DDoS. In addition, network slicing is a fundamental architecture component of the 5G/B5G. Network slicing also brings up a number of security issues from slice isolation to concurrent multiple access to slices by a single user. In this special session, we plan to address all the aforementioned issues under the big umbrella of digital convergence of 5G/B5G, AIoT and Security.

SS-28

Target track
WCN
Title
Advanced Wireless Access Technologies and Data Analysis for IoT and Environmental Monitor
Organizer(s)
Osamu Takyu
Contact info.
takyu at shinshu-u.ac.jp
Abstract
Recently, an environmental monitoring from the Internet is attracting much attention, such as Internet of Things (IoT) and Cyber Physical Systems (CPS). The measurement environment is expanded to the three-dimensional space and thus the sensing results should be gathered by wireless communication technologies. A various wireless communication system for IoT and environmental monitoring have been considered. For example, low power wide area (LPWA) networks and body area networks support the communication range of several kilometers and the short range, respectively. If the wireless access technologies are not suitable for gathering the required sensing results, the accurate environmental monitoring is missed. In this special session, for achieving the highly accurate IoT and environmental monitoring, the latest wireless access technologies are shown. We discuss the problem and the suitable solutions and thus clarify the future extension of the wireless access technologies for supporting IoT and environmental monitoring.

SS-29

Target track
MLDA
Title
Deep Learning for Few-shot Data Analysis
Organizer(s)
Xiaoxu Li, Bin Dai, Zhanyu Ma, Jen-Tzung Chien
Contact info.
xiaoxulilut at gmail.com
Abstract
Due to the rapid development of deep learning, recent years have witnessed breakthrough in various tasks with large amount of labeled samples in computer vision, natural language processing and speech signal processing. However, learning a machine learning model with good generalizability from only a few labeled samples still exists a big challenge. Furthermore, humans never use millions or billions of data when they learn knowledge, and many kinds of data are quite small in quantity in real-world, Hence, it is of great importance to focus on few-shot learning and to develop few-shot leaning methods in fields of computer vision, natural language processing and speech signal processing. In few-shot learning, the biggest issue is the deficiency of samples. Researchers have alleviated the issue from several learning types, such as meta-learning methods, transferring learning methods, metric-based methods. However, many challenging topics remain with few-shot leaning techniques, such as data generation, discriminative feature learning, metric-learning, meta-learning, etc. Moreover, overfitting, imbalance of data distribution incurred by deficiency of samples should be focused on. Therefore, the goal of this special session is to provide theoretical foundations and ground-breaking models in order to overcome the deficiency of samples for few-shot learning.

SS-30

Target track
MLDA
Title
Online and Distributed Kernel Learning Algorithms
Organizer(s)
Anthony Kuh, Toshihisa Tanaka
Contact info.
kuh at hawaii.edu
Abstract
The last several years has seen the development of online kernel algorithms combining methods from Reproducing Kernel Hilbert Spaces with principles from Adaptive Signal Processing. Considerable progress has been made in development of theory, algorithms, and applications. There is still considerable interest in this area as we consider extensions of theory and new paradigms for both algorithm development and applications. This special session addresses some of these extensions. They include kernel learning on graphs, developing distributed kernel learning algorithms, efficient learning of hyperparameters, integrating spatial and temporal kernel learning algorithms, and applications. The session will provide methods of extending kernel method theory, algorithms, and applications so that they can be more widely used in the data science and machine learning areas.