IE E E S IG N A L P R O C E S S IN G M A G A Z IN E S K E T C H IN G D A T A S E T S F O R L A R G E -S C A L E L E A R N IN G V O L U M E 3 8 N U M B E R 5 S| E P T E M B E R 2 0 2 1 Call for Papers Special Issue on Physics-Driven Machine Learning for Computational Imaging Recent years have witnessed a rapidly growing interest in next-generation imaging systems and their combination with machine learning. While model-based imaging schemes that incorporate physics-based forward models, noise models, and image priors laid the foundation in the emerging field of computational sensing and imaging, recent advances in machine learning, from large-scale optimization to building deep neural networks, are increasingly being applied in modern computational imaging. A wide range of machine learning techniques can be applied to enhance the effectiveness and efficiency of computational imaging systems, thus redefining the state-of-the-art computational imaging algorithms. On the other hand, different from image restoration problems, computational imaging involves physics-based imaging processes, physics-based properties of underlying data (e.g., characterized by partial differential equations or PDEs), and domain knowledge, which need to be exploited and combined with various data-driven approaches to benefit applications from sensing to image reconstruction. There are compelling challenges for such interdisciplinary research that remain to be addressed, ranging from learning theory to imaging applications. This special issue invites tutorial on recent works on novel imaging pipelines such as smart imaging system design, machine learning and related methods, and task and data-driven imaging system hardware and algorithm design that are tightly combined with imaging physics. Besides, this special issue will promote recent works on learning theory for computational imaging, including performance guarantees, convergence analysis, learning model analysis, etc., which are critical for reliable and interpretable computational imaging systems. Important Dates White papers due: 1 December 2021 Topics of interest include but are not limited to: Invitation notification: 1 January 2021 ● Novel Learning-Driven Computational Imaging Systems Fullmanuscripts due: 1 March 2022 ● Learning-based Modeling and Algorithms for Imaging Firstreview to authors: 1May2022 ● Learning Theory for Computational Imaging ● Computational Imaging Applications Revision due: 1 July2022 ● Ethics and Social Impacts of Learning for Computational Second review completed: 1 September, 2022 Imaging Final manuscriptdue: 1 October2022 Publication: 1 January 2023 White papers are required, and full articles will be invited based on the review of white papers. The white paper format is up to 4 pages in length, including the proposed article title, motivation and significance of the topic, an outline of the proposed paper, and representative references. An author list with contact information and short bios should also be included. Submitted articles must be of tutorial/overview/survey nature, in an accessible style to a broad audience, and have a significant relevance to the scope of the Special Issue. Submissions should not have been published or be under review elsewhere, and should be made online at https://mc.manuscriptcentral.com/sps-ieee. For submission guidelines, see the Information for Authors at https://signalprocessingsociety.org/publications-resources/ ieee-signal-processing-magazine/information- authors-spm. Guest Editors Prof. Jong Chul Ye (Lead GE),Korea Adv. Inst.of Science and Technology(KAIST), Korea, [email protected] Prof.Bihan Wen, Nanyang Technological University, Singapore, [email protected] Prof. Saiprasad Ravishankar,Michigan State University, USA, [email protected] Prof. Raja Giryes, Tel Aviv University, Israel, [email protected] Prof. Zhizhen Zhao, University of Illinois at Urbana-Champaign,mailto:[email protected] Digital Object Identifier 10.1109/MSP.2021.3094969 Contents Volume 38 | Number 5 | September 2021 FEATURES COLUMNS 12 SKETCHING DATA SETS FOR 7 Society News LARGE-SCALE LEARNING New Society Editors-in-Chief Rémi Gribonval, Antoine Chatalic, Named for 2022 Nicolas Keriven, Vincent Schellekens, Laurent Jacques, and Philip Schniter 8 Special Reports Smart Home Technologies Are 37 DIAGNOSIS/PROGNOSIS OF Saving Money and Lives COVID-19 CHEST IMAGES John Edwards VIA MACHINE LEARNING AND HYPERSIGNAL PROCESSING 86 Lecture Notes Arash Mohammadi, Yingxu Wang, Tracking and Estimation of Frequency, Nastaran Enshaei, Parnian Afshar, Amplitude, and Form Factor of a Farnoosh Naderkhani, Harmonic Time Series Anastasia Oikonomou, R onald M. Aarts Moezedin Javad Rafiee, Helder C.R. Oliveira, Svetlana Yanushkevich, and Konstantinos N. Plataniotis ON THE COVER 67 SOUND EVENT DETECTION Annamaria Mesaros, Toni Heittola, Tuomas Virtanen, The three feature articles in this issue focus on the process- and Mark D. Plumbley ing of data, especially large data in high dimension, using signal processing and machine learning methods. COVER IMAGE: ©SHUTTERSTOCK.COM/AMIAK PG. 3 PG. 8 IEEE SIGNAL PROCESSING MAGAZINE (ISSN 1053-5888) (ISPREG) is published bimonthly by the Institute of Electrical and Electronics Engineers, Inc., 3 Park Avenue, 17th Floor, New York, NY 10016-5997 USA (+1 212 419 7900). Responsibility for the contents rests upon the authors and not the IEEE, the Society, or its members. Annual member subscriptions included in Society fee. Nonmember subscriptions available upon request. Individual copies: IEEE Members US$20.00 (first copy only), nonmembers US$248 per copy. Copyright and Reprint Permissions: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy beyond the limits of U.S. Copyright Law for private use of patrons: 1) those post-1977 articles that carry a code at the bot- tom of the first page, provided the per-copy fee is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923 USA; 2) pre-1978 articles without fee. Instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. For all other copying, reprint, or republication permission, write to IEEE Service Center, 445 Hoes Lane, Piscataway, NJ 08854 USA. Copyright © 2021 by the Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Periodicals postage paid at New York, NY, and at additional mailing offices. Postmaster: Send address changes to IEEE Signal Processing Magazine, IEEE, 445 Hoes Lane, Piscataway, NJ 08854 USA. Canadian GST #125634188 Printed in the U.S.A. Digital Object Identifier 10.1109/MSP.2021.3086572 IEEEIE SEIEG SNIGALN APLR OPCREOSCSEISNSGIN MGA MGAAZGIANZEI N E| S| e pMteamy b2e0r 2210 2 1| | 11 DEPARTMENTS IEEE Signal Processing Magazine 3 From the Editor EDITOR-IN-CHIEF ASSOCIATE EDITORS—COLUMNS AND FORUM Think Outside the Box! Christian Jutten—Université Grenoble Alpes, Balázs Bank—Budapest University of Technology Christian Jutten France and Economics, Hungary Ulisses Braga-Neto—Texas A&M University, 4 President’s Message AREA EDITORS USA “She Changed Astronomy Forever. Cagatay Candan—Middle East Technical Feature Articles He Won the Nobel Prize for It.” University, Turkey Laure Blanc-Féraud—Université Côte d’Azur, Ahmed Tewfik Yuan-Hao Huang—National Tsing Hua France University, Taiwan 92 Dates Ahead Special Issues Wei Hu—Peking University, China Xiaoxiang Zhu—German Aerospace Center, Andres Kwasinski—Rochester Institute of Germany Technology, USA Columns and Forum Xin Liao—Hunan University, China Rodrigo Capobianco Guido—São Paulo Piya Pal—University of California San Diego, State University (UNESP), Brazil USA H. Vicky Zhao—Tsinghua University, Hemant Patil—Dhirubhai Ambani Institute R.P. China of Information and Communication e-Newsletter Technology, India Behnaz Ghoraani—Florida Atlantic University, Christian Ritz—University of Wollongong, USA Australia Social Media and Outreach Emil Björnson—KTH Royal Institute of Technology, ASSOCIATE EDITORS—e-NEWSLETTER Sweden Abhishek Appaji—College of Engineering, India Subhro Das—MIT-IBM Watson AI Lab, EDITORIAL BOARD IBM Research, USA Anubha Gupta—Indraprastha Institute of Information Danijela Cabric—University of Technology Delhi (IIIT), India California, USA Hamid Palangi—Microsoft Research Lab (AI), USA PG. 92 Mario Figueiredo—Instituto Superior Técnico, University of Lisbon, Portugal IEEE SIGNAL PROCESSING SOCIETY Sharon Gannot—Bar-Ilan University, Ahmed Tewfik—President Israel M/NANCY KENNEDY YJCoihfsarLeinasp tbhGino Geora nuGtgoeu—rricilelMie—sm,i cIInonroftco—s.r,o mIfUnt arSCitaAioo rRnpe oSnrynasettieso,m nF,sr aUnScAe AASKh.tnVhraC.iiSkn oIa.as nnaH fPtbeaheer rteNirl— onPapcéVreuraiselcuyze—a- NPnPraeerniesri—sadi—deVnePVtn, itEMc-Edeleue Pmccratebtseioirdsnehinpt, ©SHUTTERSTOCK.CO NEMPariaaktFrg riLriceaaaker n sGnLcse oe.H nCMi—maalLalrietnyitnka—iöt——pUiKnnIngiivnt eeUglr,sns tiUiovtyneS r AosUift nyNi,v aSewnrsteeitdsy,,e nU.K. MTIEüEalaEr ycS IAMGdNoaoAlnıL—e PnVR—iOcVeC iEPcSerSe PIsNirdeGes inSdtO,e nCTetI,Ec TPhuYnbi cSliaTcAla DFtiFoirnesc tions The IEEE Workshop on Applications of Signal Processing William Colacchio—Senior Manager, Publications Helen Meng—The Chinese University to Audio and Acoustics will be held 17–20 October in New and Education Strategy and Services Paltz, New York, United States. of Hong Kong, China Rebecca Wollman—Publications Administrator Meinard Mueller—Friedrich-Alexander Universität Erlangen-Nürnberg, Germany Phillip A. Regalia—U.S. National Science Foundation, USA IEEE PERIODICALS MAGAZINES DEPARTMENT Alejandro Ribeiro—University of Jessica Welsh, Managing Editor Pennsylvania, USA Geraldine Krolin-Taylor, Douglas O’Shaughnessy—INRS Universite de Senior Managing Editor Recherche, Canada Janet Dudar, Senior Art Director Osvaldo Simeone—Kings College London, Gail A. Schnitzer, Associate Art Director U.K. Theresa L. Smith, Production Coordinator Milica Stojanovic—Northeastern University, Mark David, Director, Business Development - USA Media & Advertising Dimitri Van De Ville—Ecole Polytechnique Felicia Spagnoli, Advertising Production Manager Fédérale de Lausanne, Switzerland Jong Chul Ye—KAIST, South Korea Peter M. Tuohy, Production Director Josiane Zerubia—INRIA Sophia-Antipolis Kevin Lisankie, Editorial Services Director Mediterranee, France Dawn M. Melley, Senior Director, Qing Zhao—Cornell University, USA Publishing Operations Digital Object Identifier 10.1109/MSP.2021.3086573 SCOPE: IEEE Signal Processing Magazine publishes tutorial-style articles on signal processing research and Promoting Sustainable Forestry applications as well as columns and forums on issues of interest. Its coverage ranges from fundamental principles to practical implementation, reflecting the multidimensional facets of interests and concerns of the community. Its SFI-01681 mission is to bring up-to-date, emerging, and active technical developments, issues, and events to the research, IEEE prohibits discrimination, harassment, and bullying. educational, and professional communities. It is also the main Society communication platform addressing important For more information, visit issues concerning all members. http://www.ieee.org/web/aboutus/whatis/policies/p9-26.html. 2 IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | FROM THE EDITOR Christian Jutten | Editor-in-Chief | [email protected] Think Outside the Box! Summer is a nice, quiet period, with manifold with suited non-Euclidian the end of the academic year in many metrics [4]. Finally, in the mid-1990s, countries. It is a time for vacations he focused on modeling retina, first and to think about events and develop a the neural layers after the retina and vision for our future. Of course, this vi- more generally vertebrate visual percep- sion is based on current research results, tion, for understanding how motions, biased by fashion effects and publish or textures, colors, etc., are processed in perish tendencies and perturbated by the the brain. These works result in many COVID-19 pandemic, which, in depth, powerful bioinspired algorithms, as modifies our life—professional as well well as applications and patents, which as personal. are summarized in his very inspir- First, I would like to evoke the mem- ing book Vision: Images, Signals and ory of Prof. Jeanny Hérault, who died Neural Networks. Models of Neural on 7 June 2021. Since 1970, Jeanny had Prof. Jeanny Hérault, keynote speaker at the Processing in Visual Perception [5] IEEE MLSP 2009 workshop, in Grenoble, France. been a faculty member at the Polytech- published in 2010. Finally, just after he nic Institute and then at the University retired, he created a nonprofit associa- Joseph Fourier in Grenoble, France, innovative methods in signal and im- tion in Grenoble, open to any curious and he trained dozens of Ph.D. stu- age processing always inspired by person, called Neurocercle (https:// dents, postdocs, and young research- brain and human vision, including ar- neurocercle.wordpress.com), and each ers. He was my Ph.D. supervisor before tificial neural networks. In early 1980, month since 2009, he has invited a becoming my friend, and I am sure that inspired by modeling how the brain scientist to talk about advances in neu- most of you understand from your own of vertebrates can decode arm and leg rosciences and cognitive sciences and experiences how strong our friend- motions, he proposed the basics of implications in our lives. ships could be. Some of you knew him blind source separation and indepen- as well as his contributions: he was dent component analysis [1]. In the In this issue actually an out-of-the-box scientist. In 1990s, since computers were not pow- Three feature articles are the core of this this editorial, I will recall some of his erful enough to run artificial neural issue of IEEE Signal Processing Maga- outstanding contributions around sig- networks, Carver Mead [2] in the Unit- zine (SPM). They are focused on the pro- nal processing. ed States and a few scientists around cessing of data, especially large data in Very early on, Jeanny was fascinated the world developed, like Jeanny and high dimension, using signal processing by the brain, as it is a fantastic machine several of his Ph.D. students, a few and machine learning methods. Clearly, for signal processing. When he decided special architectures able to train and in these articles, an essential ingredient to prepare a Ph.D. in 1968, he chose a efficiently use neural networks such is the data. The usual way is to process topic fully out of the box focused on as multilayer perceptrons and self-or- always larger data sets, larger high-di- modeling neural cells and simulating ganizing maps [3]. He then developed mension data with very powerful, mul- them on electronic circuits. And then, the concept of curvilinear component ticore computers. The articles point out during his entire career, he proposed analysis, which extends the Kohonen’s different tricky learning issues in data maps to huge-dimension data and sets, i.e., how to deal with unbalanced implicitly considers high-dimension Digital Object Identifier 10.1109/MSP.2021.3086574 Date of current version: 27 August 2021 data are embedded in a low-dimension (continued on page 5) IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | 3 PRESIDENT’S MESSAGE Ahmed Tewfik | IEEE Signal Processing Society President | [email protected] “She Changed Astronomy Forever. He Won the Nobel Prize for It.” It’s been a while since I last wrote a col- nical, geographical, and gender diver- representatives on publications, con- umn for IEEE Signal Processing Mag- sity. I won’t belabor again the benefits ferences, and boards; organizing com- azine. I will try to address here some of diversity. Nor will I detail the real mittees; publication and conferences of the many questions and changes systemic hurdles that more than half of boards; executive committee mem- that arose since the beginning of the mankind experiences within society, at bers; president-elect slates; board of year. But before I do so, I would like to work, home, or our own technical Soci- governors members; and all staff posi- invite you to watch a short documen- ety. We are all very familiar with these tions. Without a strong and diverse pool tary by Ben Proudfoot with the exact facts and may even be convinced that of candidates, we won’t have access to title of this column: “She Changed As- we understand them. Perhaps we do. the best candidates for any given role, tronomy Forever. He Won the Nobel But it appears to me that, in many cas- will come short on innovation, and Prize for It.” This story is yet another es, this understanding is abstract rather won’t dismantle the barriers that hold reminder of how diversity is sine qua than visceral. Hence, it doesn’t reflect many of us back. non for innovation. The documentary in our actions. Another topic that keeps coming up covers the discovery of pulsars in 1967 Our goals as a Society include achiev- is the relationship between signal pro- by Jocelyn Bell Burnell. Her discov- ing excellence and fairness, empower- cessing and machine learning. Many of ery changed our view of the universe. ing all our members, and helping our us appear frustrated that our conferenc- Dr. Bell was a Ph.D. student at the time, fellow men and women achieve their es and publications aren’t garnering the studying under renowned astronomer full potential. Quotas won’t get us there same attention and prestige as machine Anthony Hewish. She overcame nu- and can be damaging. Online training learning publications and conference merous barriers as the only woman in and statements that profess our belief venues. Others bemoan the fact that our many classes and in the Ph.D. program. in equity and diversity aren’t enough. best students prefer to submit their best Her persistence and unconventional I personally believe that it is our indi- work to machine learning conferences thinking led her to uncover pulsars vidual and collective responsibility to rather than ICASSP or ICIP. Machine when others saw only normal electro- source diversity. Results are what mat- learning is yet another arrow in our magnetic interference. ters. It is our responsibility to find the quiver, no different than linear alge- Despite the remarkable progress we underrepresented talent that espouse bra, probability and statistics, number have made since Dr. Bell was a student, unconventional thinking and creativity theory, information theory, and the host lack of diversity remains one of our most and take calculated risks, regardless of of other disciplines that we leverage to pernicious and frustrating challenges in their external appearance or extrinsic bring magic to humankind. It may ac- 2021. Diversity is an issue that we all label arbitrarily applied to them by so- tually be a better arrow, particularly profess to believe in. However, our ac- ciety, to foster innovation and be true to when judiciously combined with other tions often belie our declared values. the aforementioned goals. That is why innovations that we as a community Time and again this year, we were sent we, as a Society, passed a resolution that contributed to human knowledge. We nomination slates for various positions, mandates the sourcing of diversity in all don’t seem to be bothered by the fact from representatives of the Society on our functions. We expect all our mem- that core linear algebra meetings, pub- cosponsored conferences to executive bers, committees, and boards to actively lications, and Societies are where most committee members that lacked tech- seek diversity in generating candidates core linear algebra results are presented for all volunteer and staff positions, from and discussed. Why is the situation dif- Distinguished Lecturers to panelists; ferent when it comes to machine learn- Digital Object Identifier 10.1109/MSP.2021.3100655 Date of current version: 27 August 2021 invited speakers to keynote speakers; ing? Indeed, our computer science 4 IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | colleagues have leveraged signal pro- town halls will give you an opportunity presentation in virtual meetings. Such cessing tools to advance machine learn- to ask candidates direct questions about an approach has the advantage of quick ing. No one has exclusive ownership issues that are dear to you or suggest to validation and interactive dissemina- of any intellectual construct or tool. the candidates the initiatives that will tion of innovation without sacrificing Transformational innovations that re- benefit all of us. The webinars will fo- review quality, in contrast to the current sult from creative insights and cross cus on strategic planning, as I indicated latency built in our journal and even disciplinary synergies deservedly get in a previous column. I urge you to bring conference publication mechanisms. the limelight. That is precisely what we innovative ideas to these webinars. For Should we explore such unconventional should focus on. example, building on the lessons we’ve publication fora? To foster innovation and unconven- learned from virtual conferences in the tional thinking, we’re also inviting you COVID-19 era, some of our computer to participate in a number of online we- science colleagues are promoting the binars and town halls. New this year are idea of periodic paper submission op- town halls for you to meet candidates portunities, followed by a fast and rig- for the president-elect position. These orous review process and immediate SP FROM THE EDITOR (continued from page 3) classes, with missing data, with too- of compressive sensing to compressive dare to investigate these a little crazy small data sets, etc. The articles suggest learning, i.e., to extract a small num- and out-of the-box ideas! smart solutions for overcoming these ber of lower-dimension features based problems, e.g., how one can artificially on averaged nonlinear mappings of the generate additional relevant data for initial data. This idea presents many completing missing or unbalanced data, advantages: it avoids storing huge or how one can use transfer learning for data in memory; it leads to much faster reducing learning time and complexity. learning algorithms, able to run on sim- References Thus, based on even larger completed ple embedded devices; it can achieve [1] C. Jutten and J. Hérault, “Blind separation of sources, part I: An adaptive algorithm based on neuro- data sets, requiring higher computation- similar or even better performance than mimetic architecture,” Signal Process., vol. 24, no. 1, al loads, performance can be improved. using the complete data sets. Sketching 1991, pp. 1–10. doi: 10.1016/0165-1684(91)90079-X. However, there is another unusual is also able to preserve data privacy, [2] C. Mead and M. Ismail, Analog VLSI implemen- tation of neural systems, (The Kluwer International way to address the scale problem of which is very important when using Series in Engineering and Computer Science). Norwell, processing huge data sets, an out-of- sensitive data. MA: Kluwer Academic Publishers, 1989, vol. 80. [3] J. C. Lawson, N. Maria and J. Hérault, “SMART: the-box idea perfectly illustrated by Of course, it is much more com- neurocomputer using sparse matrices,” in Proc. the first feature article “Sketching Data fortable to do incremental research, Euromicro Workshop on Parallel Distrib. Process., 1993, pp. 59–64. doi: 10.1109/EMPDP.1993.336420. Sets for Large-Scale Learning” in this following the current flow of research. [4] P. Demartines and J. Hérault, “Curvilinear com- issue of SPM. Instead of enlarging the Sure, it is risky to be out of the box: ponent analysis: A self-organizing neural network for data sets, the basic idea of sketching it requires a lot of effort to explain nonlinear mapping of data sets,” IEEE Trans. Neural Netw., vol. 8, no. 1, pp. 148–154, 1997. doi: is to only keep the data you need. The new ideas, and it also will take longer 10.1109/72.554199. idea of sketching is particularly intelli- and be more difficult to be published. [5] J. Hérault, “Vision: Images, signals and neural gent as well as ecological. Faced with But, I believe that such out-of-the-box networks,” in Models of Neural Processing in Visual Perception. World Scientific, 2010. doi: huge data learning, the main principle ideas are actually mandatory for in- 10.1142/7311. of sketching is to extend the concept novation and science advances. So, SP IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | 5 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing human-centric signal processing May 22 - 27, 2022, Sands Expo & Convention Centre https://2022.ieeeicassp.org/ General Chair The International Conference on Acoustics, Speech, & Signal Processing (ICASSP), is the IEEE Signal Haizhou Li (NUS, Singapore) Processing Society’s flagship conference on signal processing and its applications. The 47th edition of ICASSP will be held in Singapore. The programme will include keynotes by pre-eminent international General Co-Chair Sadaoki Furui (NII, Japan) speakers, cutting-edge tutorial topics, and forward-looking special sessions. ICASSP also provides a great networking opportunity with a wide range of like-minded professionals from academia. Honorary Chair Tsuhan Chen (NUS, Singapore) CONFERENCE TOPICS Technical Program Chairs • Applied Signal Processing Systems • Multimedia Signal Processing Woon-Seng Gan (NTU, Singapore) • Audio & Acoustic Signal Processing • Quantum Signal Processing Kai-Kuang Ma (NTU, Singapore) • Bio Imaging & Signal Processing • Remote Sensing and Signal Finance Chairs • Compressive Sensing, Sampling Processing Lei Wang (COLIPS, Singapore) & Dictionary Learning • Sensor Array & Multichannel Yan Wu (I2R, A*STAR, Singapore) • Computational Imaging Signal Processing Special Session Chairs • Computer Vision Financial Signal Processing • Signal Processing for Big Data Jonathon Chambers (Uleicester, UK) • Deep Learning/Machine Learning • Signal Processing for Communication Zhengguo Li (I2R, A*STAR, Singapore) for Signal Processing & Networking Pina Marziliano (EPFL, Switzerland) • Image, Video & Multidimensional • Signal Processing for Cyber Security Plenary Chairs Signal Processing • Signal Processing for Education C.-C. Jay Kuo (USC, USA) • Industrial Signal Processing • Signal Processing Over Graphs Helen Meng (CUHK, HK SAR, China) • Information Forensics & Security • Signal Processing Theory & Methods Douglas O’Shaughnessy (INRS, Canada) • Internet of Things • Speech & Language Processing Tutorial Chairs Mauro Barni (USiena, Italy) Call for Tutorials:Tutorials at ICASSP form an important part of the program, giving attendees the Saeid Sanei (NottinghamTU, UK) opportunity to learn about current research areas that are of growing interest to the signal processing Tanja Schultz (U Bremen, Germany) community. Therefore, the tutorials should be prepared and presented to a wide audience. Prospective Publication Chairs presenters may discuss their proposals with the tutorials chair before the submission. Kenneth Lam (HKPU, HK SAR, China) Berrak Sisman (SUTD, Singapore) Yan Sun (URI, USA) Call for Special Sessions:The program for ICASSP 2022 will include Special Sessions that complement the main program, in line with the technical scope of the conference. Please refer to the Local Arrangement Chairs conference webpage for information about Special Session proposals and submissions. Minghui Dong (I2R, A*STAR, Singapore) Kong Aik Lee (I2R, A*STAR, Singapore) Siqi Cai (NUS, Singapore) Call for Exhibitors and Sponsors:ICASSP offers exhibitors and companies an opportunity to showcase their products and innovative solutions, as well as recruitment and networking opportunities. Show and Tell Chairs Please check the conference webpage for information about signing up to become an exhibitor or Nancy Chen (I2R, A*STAR, Singapore) Tomi Kinnunen (UEF, Finland) sponsor at ICASSP 2022. Rong Tong (COLIPS, Singapore) SP Society Journal Paper Presentations:Authors of papers in the journals of the IEEE Signal Student and Education Activities Chairs Rongshan Yu (XMU, China) Processing Society will have the opportunity to present their work at ICASSP 2022, subject to space Kenan Emir Ak (I2R, A*STAR, Singapore) availability and approval by the Technical Program Chairs. Only papers published between October 20, 2020 and October 19, 2021 are eligible for presentation at ICASSP 2022. The papers presented at ICASSP Sponsorship Chairs will neither be reviewed nor included in the proceedings for the IEEE Xplore Digital Library. Zhizheng Wu (Facebook, USA) Lei Xie (NWPU, China) Open Preview:Open Preview allows conference proceedings to be available in the IEEE Xplore Publicity Chair Digital Library, free of charge, to all customers, 30 days prior to the conference start date, through the Eliathamby Ambikairajah (UNSW, Australia) conference end date. Grand Challenge Chairs Andy Khong (NTU, Singapore) IMPORTANT DATES Wee Peng Tay (NTU, Singapore) Industry & Exhibits Chairs Special Session Proposal ................................................. 9 August 2021 Fa-Long Luo (Micron, USA) Special Session Acceptance .................................... 6 September 2021 Wei Lu (SUTD, Singapore) Tutorial Proposals ............................................................ 11 October 2021 Soujanya Poria (SUTD, Singapore) Paper Submissions ............................................................. 1 October 2021 Women’s Activities Chair Grand Challenge Proposal ............................................ 18 October 2021 Dorien Herremans (SUTD, Singapore) Grand Challenge Acceptance .................................... 8 November 2021 Tutorial Acceptance ................................................... 15 November 2021 Social Media Chair Chitralekha Gupta (NUS, Singapore) SPS Journal/Letters Due ................................................ 10 January 2022 Paper Acceptance ........................................................... 21 January 2022 American Liaison Final Paper ........................................................................ 17 February 2022 Jane Wang (UBC, Canada) Author Registration ........................................................ 17 February2022 Roxana Saint-Norm (UADE, Argentina) Open Preview ......................................................... 22 April - 27 May 2022 Singapore Chapter Liaison Show and Tell Proposals ..................................................... 7 March 2022 Ngai-Man (Man) Cheung (SUTD, Singapore) Digital Object Identifier 10.1109/MSP.2021.3102794 SOCIETY NEWS New Society Editors-in-Chief Named for 2022 The following volunteers have been He is a Fellow of IEEE and is with the University of Padua, Italy. He is suc- named editors-in-chief of IEEE Sig- University of Illinois at Urbana–Cham- ceeding Patrizio Campisi, University of nal Processing Society publications. paign. He is succeeding Dilek Hakkani- Roma TRE, who has held the post The term for these editors-in-chief will Tur, Amazon Alexa AI, who has held since 2018. run from 1 January 2022 through 31 the post since 2019. December 2024. Pascal Frossard Mujdat Cetin Pascal Frossard is the Xiao-Ping Zhang Mujdat Cetin is the new editor-in-chief of Xiao-Ping Zhang has new editor-in-chief of IEEE Transactions on been named editor-in- IEEE Transactions on Signal and Informa- chief of IEEE Journal Computational Imag- tion Processing Over of Selected Topics in ing. He is a Fellow of Networks. He is a Signal Processing. He IEEE and is with the Fellow of IEEE and is with EPFL, is a Fellow of IEEE University of Rochester, New York. He Switzerland. He is succeeding Anto- and is with Ryerson University, Cana- is succeeding Brendt Wohlberg, Los nio Ortega, University of Southern da. He is succeeding Lina J. Karam, Alamos National Laboratory, who has California, who has held the post Lebanese American University, who been in the role since 2018. since 2019. has held the post since 2019. Mauro Conti SP Paris Smaragdis Mauro Conti has taken Paris Smaragdis has the role of editor-in- taken the role of edi- chief of IEEE Trans- tor-in-chief of IEEE/ actions on Information ACM Transactions on Forensics and Securi- Audio, Speech, and ty. He is a Senior Language Processing. Member of IEEE and is with the Digital Object Identifier 10.1109/MSP.2021.3094863 Date of current version: 27 August 2021 IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | 7 SPECIAL REPORTS John Edwards Smart Home Technologies Are Saving Money and Lives Reaching out in new directions, signal processing-supported smart technologies are rapidly changing—and improving—everyday life Smart home technologies, designed also a deep learning network that mod- tions created by multiple appliances by to make users happier, healthier, els vibration data to create different sig- detecting the distinct noise each device and wealthier, are rapidly becom- natures for each appliance. creates as well as by identifying the paths ing a mainstay of everyday life. In most Many home appliances generate traveled by each vibration as it moves cases, signal processing is essential to the vibrations identifiable by unique signa- from room to room. VirbroSense’s deep devices’ operation and performance. tures that propagate through a home’s learning network is trained to distin- These days, a variety of intelligent walls, ceilings, and other infrastructure guish from among different activities. automated devices can be found in elements, observes Cheng Zhang, a Cor- In tests spanning five separate nearly every home. The trend is accel- nell assistant professor of computing and houses, VibroSense achieved a nearly erating so rapidly that it now appears information science as well as director of 96% accuracy, identifying 17 different inevitable that smart technology will the university’s SciFi Lab. “These unique activities, including dripping faucets, an soon be integrated into virtually every characteristics … can be captured by a exhaust fan, an electric kettle, a refrigera- facet of daily life. laser vibrometer and recognized by AI tor, and a range hood (see Figure 1). The With smart technologies already (artificial intelligence) technology—in technology was also able to distinguish widely used in applications as diverse our case, a deep learning algorithm—to five different stages of appliance usage as home security, environmental con- distinguish the sound source.” with an average accuracy of more than trol, and home entertainment, research- Zhang says that VibroSense promises 97%. In single-story houses, the laser ers are now heading in new directions, to allow users to monitor multiple house- was pointed at an interior wall at the cen- leading to applications as diverse as hold appliance activities with just a single ter of the home. The device was pointed home appliance monitoring, improved sensor. Besides alerting users to runaway at the ceiling in two-story homes. power delivery, and advanced remote appliances, the technology also promises Signal processing was the first tech- medical support. to give homeowners a way to track ener- nology the researchers considered when gy expenses. “Knowing which device exploring the system’s feasibility. “We Good vibrations is working, and for how long and under use signal processing to help us ana- For anyone who has ever neglected to which mode, would potentially help the lyze the information from the captured turn off an electric kettle, water faucet, owner of the house to estimate and better vibration data,” Zhang says. Various air conditioner, or other household appli- manage power usage,” he notes. types of filters are then used to remove ance before leaving home—potentially To create a smart home using con- noise. “We visualize the signal in both leading to unfortunate consequences— ventional technology, it’s currently time and frequency domains, which Cornell University researchers are offer- necessary to equip each appliance with helps us to find any distinguishable ing a potential safeguard. The team’s smart technology, Zhang observes. “Our features,” he notes. A Fourier trans- new VibroSense device is a smart home system is the first that can monitor devic- form is used to convert the data from technology that can track 17 types es across different floors, in different time to frequency domain. “We also of appliances using subtle vibrations rooms, using a single device.” use wavelet transformation and band- found in walls, ceilings, and floors; it is To monitor appliance usage through- pass filter to help remove the noise and out an entire house, VibroSense is improve the signal-to-noise ratio.” equipped with a laser Doppler vibrom- Signal processing methods also enab- Digital Object Identifier 10.1109/MSP.2021.3088401 Date of current version: 27 August 2021 eter. The device differentiates the vibra- led the researchers to quickly verify an 8 IEEE SIGNAL PROCESSING MAGAZINE | September 2021 | 1053-5888/21©2021IEEE