Keynote Speaker

Prof. Antanas Verikas
Halmstad University, Sweden

Prof. Antanas Verikas was awarded a PhD degree in pattern recognition from Kaunas University of Technology, Lithuania. Currently he holds a professor position at both Halmstad University Sweden, where he leads the Department of Intelligent Systems, and Kaunas University of Technology, Lithuania. His research interests include learning systems, classification, fuzzy logic, image processing, computer vision, pattern recognition, applied soft computing, and visual media technology. He published more than 170 peer reviewed articles in international journals and conference proceedings and served as Program committee member in numerous international conferences. He is a member of the European Neural Network Society, International Pattern Recognition Society, International Association of Science and Technology for Development, and a member of the IEEE.

Speech Title: Modelling Speech Signals for Parkinson's Disease Screening

Abstract: Parkinson's disease (PD) is the second most common neurodegenerative disease after Alzheimer's and it is anticipated that the prevalence of PD is going to increase due to population ageing. This study investigated sustained phonation and text-dependent speech modalities for Parkinson's disease screening. Signals were recorded through two channels simultaneously, namely, acoustic cardioid (AC) and smart phone (SP) microphones.

Information in each modality was summarized by 18 well-known audio feature sets. The sustained phonation modality was also explored by applying signal decomposition into intrinsic mode functions (IMFs), namely, the empirical mode decomposition (EMD) and the variational mode decomposition (VMD). Random forest (RF) was used as a machine learning algorithm, for both individual feature sets and for decision-level fusion. Non-linear projection of an RF-based proximity matrix into the 2D space enriched medical decision support by visualization.

The voice signal decomposition into IMFs followed by the decision-level fusion was capable of providing excellent detection performance. The out-of-bag equal error rate (EER) of ~1% for the AC and ~12% for the SP channel was observed. Application of convolutional neural networks (CNN) on text-dependent speech recordings resulted in the EER of 14.1% for the AC channel. Besides the common Mel-frequency spectrogram and its first and second derivatives, various other input feature maps were also used in the CNN.

Detection performance was consistently better for the AC than for SP microphone. Nonetheless, sustained phonation and/or text-dependent speech recordings of SP quality have potential for PD detection. Additional information is worth considering by tracking an accelerometer signal, for example. Drawing an Archimedean spiral is an interesting type of tactile task which could be performed using a hand-held device. Fusion of information from diverse non-invasive modalities could help to develop an efficient SP-based tool for PD screening.

Plenary Speaker

Assoc. Prof. Brad Mehlenbacher
Department of Educational Leadership, Policy, and Human Development (ELPHD), College of Education, NC State University, USA

Dr. Brad Mehlenbacher is currently a Visiting Scholar at the University of Waterloo's Games Institute. He is an Associate Professor of Distance Learning (Educational Leadership, Policy, and Human Development), Primary Area Faculty Member with Human Factors and Applied Cognition (Psychology), Affiliated Faculty Member with Communication, Rhetoric, and Digital Media (English and Communication), and Affiliated Faculty Member with the Digital Games Research Center (Computer Science) at NC State University. Mehlenbacher is author of the CCCC's 2012 Best Book in Technical and Scientific Communication, Instruction and Technology: Designs for Everyday Learning (MIT Press, 2010), co-author of Online Help: Design and Evaluation (Ablex, 1993), and has chapters in the CCCC award-winning Solving Problems in Technical Communication (U of Chicago Press), The Human-Computer Interaction Handbook (Lawrence Erlbaum), The Computer Science and Engineering Handbook (CRC), and the 1998 NCTE award-winning Computers and Technical Communication (Ablex). He earned his BA and MA at the University of Waterloo and his PhD in Rhetoric at Carnegie Mellon University. Mehlenbacher is past president of ACM SIGDOC. Brad has consulted for the Computer Science Department and Engineering Design Research Center at Carnegie Mellon; the Centre for Professional Writing at the University of Waterloo; Apple Computer; SAS Institute; and IBM.

Prof. Xabier Basogain
University of the Basque Country, Bilbao, Spain

Xabier Basogain is professor of the University of the Basque Country - Euskal Herriko Unibertsitatea. He is doctor engineer of telecommunications by the Polytechnic University of Madrid, and member of the Department of Engineering Systems and Automatics of the School of Engineering of Bilbao, Spain. He has taught courses in digital systems, microprocessors, digital control, modeling and simulation of discrete events, machine learning, and collaborative tools in education. His research activities include the areas of: a) soft computing and cognitive sciences to STEM; b) learning and teaching technologies applied to online education and inclusive education; c) augmented and virtual reality with mobile technologies.