Advanced Technologies and Societal Change

Reiner WichertHelmut Klausing Editors

Ambient Assisted Living7. AAL-Kongress 2014 Berlin, Germany, January 21–22, 2014

Advanced Technologies and Societal Change

More information about this series at http://www.springer.com/series/10038

Reiner Wichert • Helmut KlausingEditors

Ambient Assisted Living7. AAL-Kongress 2014 Berlin, Germany,January 21–22, 2014

123

EditorsReiner WichertFraunhofer-Allianz AAL/Fraunhofer IGDDarmstadtGermany

Helmut KlausingVDE-VerbandsgeschäftsstelleVerband der Elektrotechnik ElektronikInformationstechnik e.V.

FrankfurtGermany

ISSN 2191-6853 ISSN 2191-6861 (electronic)Advanced Technologies and Societal ChangeISBN 978-3-319-11865-9 ISBN 978-3-319-11866-6 (eBook)DOI 10.1007/978-3-319-11866-6

Library of Congress Control Number: 2013954025

Springer Cham Heidelberg New York Dordrecht London© Springer International Publishing Switzerland 2015This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or partof the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmissionor information storage and retrieval, electronic adaptation, computer software, or by similar ordissimilar methodology now known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in thispublication does not imply, even in the absence of a specific statement, that such names are exemptfrom the relevant protective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in thisbook are believed to be true and accurate at the date of publication. Neither the publisher nor theauthors or the editors give a warranty, express or implied, with respect to the material containedherein or for any errors or omissions that may have been made.

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media(www.springer.com)

Preface

Ambient Assisted Living is a research area with the potential for significant eco-nomic and social impact. While this potential has been recognized for some time,breakthroughs in terms of widespread availability and deployment of solutions areyet to be achieved. The EU and the AAL Association have funded activities in thisarea for some years, and some of these are now at a stage in their developmentwhere direct hands-on involvement of development companies is the best way tomake sure that this work produces results that are effective and applicable in realindustrial settings.

For this reason a conference series has been established as an annual showcaseevent for the people involved in this community: the AAL-Kongress (Congress forAmbient Assisted Living) with its purpose to exhibit and demonstrate ICT solu-tions, promote networking within the community, provoke debate on various topicsand highlight new or emerging developments in the area to inform the AALcommunity and discuss the problems and challenges we have to face in the com-mon years. The first AAL Kongress 2008 had the focus on applications of intel-ligent assistive systems within the areas of “health and homecare”, “safety andprivacy”, “maintenance and housework” and “social environment”. At the secondAAL-Kongress more than 520 participants attended. It focused on use cases tosupport the manufacturing of products adjusted to the needs of the user. In 2010, thethird AAL-Kongress was organized with close to 600 participants also with thefocus on use cases. In 2011 it advanced to the leading congress for AAL with 870participants. In 2012 the focus was laid on technologies in a self-determined life andthe number of participants passed over 1,000, still addressing economic challengesand trendsetting applications on innovative technology. In 2013 the sixth AAL-Kongress was focused on “quality of life in times of changing demography andtechnology”.

This 7th German AAL Congress in 2014 was entitled “Housing—Care—Socialparticipation”. It provided an excellent platform for the innovative field of AmbientAssisted Living and allowed a qualified information and knowledge exchangebetween researchers and developers, manufacturers and users, service providers,end users and representatives from politics, industry and associations. The focus

v

had been on technical solutions and concept studies from tomorrow to the nextdecade.

Within the thematic topic of this conference “Better Life with Assistive Tech-nologies” the basic human needs had been addressed in the different areas ofhousing, mobility, work, health and care. Intelligent assistance systems help peopleto lead an independent life. Innovative applications have been developed for all lifestages and situations and brought to market in recent years.

From the large number of contributions from the call for papers a selection wasmade with topics such as: Interaction and Robotics, Assistive Technologies,Monitoring and Interoperability. To underline the research priority the researchpapers have been evaluated more restrictively. More than 170 papers have beensubmitted to the seventh AAL-Kongress. After a solid review process nine papersand six posters were accepted to be included in these scientific proceedings of theconference. Three independent reviewers were matched by their expertise area tothe topic of each paper.

In closing, I would like to thank the reviewers of the Reviewing Committee, theorganizers of this event and all of the paper presenters and conference participantswho helped to make AAL-Kongress 2014 a success.

Program Co-chair for Technical Research PapersReiner Wichert (Fraunhofer-Allianz AAL/Fraunhofer IGD)

vi Preface

Program Committee AAL Kongress 2014

Axel Viehweger Verband Sächsischer Wohnungsgenossenschaften e.V.,Dresden

Uwe Fachinger Universität Vechta (stellv.)Udo Gaden Sozialwerk St. Georg e.V., Gelsenkirchen (stellv.)Sibylle Meyer SIBIS Institut für Sozialforschung und Projektberatung

GmbH, BerlinReiner Wichert Fraunhofer IGD, Darmstadt (stellv.)Jan Alexandersson DFKI, SaarbrückenMartin Braecklein Robert Bosch Healthcare GmbH, StuttgartMatthias Brucke embeteco GmbH and Co. KG, OldenburgBernd Dechert ZVEH, FrankfurtWolfgang Deiters Fraunhofer ISST, DortmundBirgid Eberhardt Tellur Gesellschaft für Telekommunikation mbH, StuttgartMelina Frenken OFFIS, OldenburgPetra Friedrich Hochschule KemptenSabine Häring Miele and Cie. KG, GüterslohArmin Hartmann Hartmann Real Estate, BochumAndreas Hein Universität OldenburgStefan Heusinger DKE, FrankfurtAnnette Hoppe Locate Solution GmbHBenno Kotterba md-pro GmbH, KarlsruheHarald Klaus Deutsche Telekom AGHarald Künemund Universität VechtaJoachim Latt Bosch Sicherheitssysteme GmbH, KasselHeidrun Mollenkopf BAGSO e.V., Expertenrat/Demenz Support, StuttgartAsarnusch Rashid FZI KarlsruheCord Schlötelburg DGBMT, FrankfurtGudrun Stockmanns Hochschule NiederrheinClaus Wedemeier GdW Bundesverband deutscher Wohnungs- und

Immobilienunternehmen e.V.

vii

Christine Weiß VDI/VDE Innovation + Technik GmbH, BerlinRalph Welge Leuphana Universität LüneburgVolker Wittpahl Ingenieurs- und Innovationsbüro, OldenburgAnton Zahneisen SOPHIA Consulting and Concept GmbH, Bamberg

viii Program Committee AAL Kongress 2014

Contents

Part I Interaction and Robotics

Challenges in Adopting Speech Control for Assistive Robots . . . . . . . . 3Paul Panek and Peter Mayer

Design of the Human-Robot Interaction for a Semi-AutonomousService Robot to Assist Elderly People . . . . . . . . . . . . . . . . . . . . . . . . 15Marcus Mast, Michael Burmester, Birgit Graf, Florian Weisshardt,Georg Arbeiter, Michal Španěl, Zdeněk Materna, Pavel Smržand Gernot Kronreif

Part II Assistive Technologies

Indoor and Outdoor Mobility Assistance . . . . . . . . . . . . . . . . . . . . . . 33Bernd Krieg-Brückner, Christian Mandel, Christoph Budelmann,Bernd Gersdorf and Antonio B. Martínez

NeuroCare—Personalization and Adaptationof Digital Training Programs for Mild Cognitive Impairments. . . . . . . 53Sandro Hardy, Christian Reuter, Stefan Göbel, Ralf Steinmetz,Gisa Baller, Elke Kalbe, Abdelkarim El Moussaoui, Sven Abels,Susanne Dienst, Mareike Dornhöfer and Madjid Fathi

Part III Monitoring

Monitoring of Therapeutic Progress by COMES® . . . . . . . . . . . . . . . . 67T. Spittler, D. Polterauer, J. Clauss, P. Friedrich and B. Wolf

ix

Part IV Interoperabilität

Information Logistics Solutions to Cope with Big DataChallenges in AAL and Telemedicine . . . . . . . . . . . . . . . . . . . . . . . . . 77Sven Meister and Wolfgang Deiters

AAL-Onto: A Formal Representation of RAALIIntegration Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89Ralph Welge, Björn-Helge Busch, Klaus Kabitzsch,Janina Laurila-Dürsch, Stefan Heusinger, Myriam Lipprandt,Marco Eichelberg, Elke Eichenberg, Heike Engelien, Murat Gök,Guido Moritz, Andreas Hein and Tim Dutz

Part V Posters

User Integration by the Evaluation of an EmergencyCall System in the Context of the Research Project MOBECS . . . . . . . 105Simon Timmermanns, Andreas Felscher, Anna Heindorf,Frank Wallhoff and Markus Meiss

Which AAL Middleware Matches My Requirements?An Analysis of Current Middleware Systemsand a Framework for Decision-Support . . . . . . . . . . . . . . . . . . . . . . . 111Tom Zentek, Can Oliver Yumusak, Christian Reicheltand Asarnusch Rashid

Human Body Detection Using Redundant RadioSignal Strength Readings for Reliable Device-Free Localization . . . . . . 127Andreas Fink, Johannes Lange and Helmut Beikirch

The WieDAS AAL Platform: Architecture and Evaluation . . . . . . . . . 139Reinhold Kröger, Wolfgang Lux, Ulrich Schaarschmidt,Jan Schäfer, Marcus Thoss, Oliver von Fragstein and Mirco Kern

Detecting the Effect of Alzheimer’s Diseaseon Everyday Motion Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149Thomas Kirste, Philipp Koldrack, Susanne Schubert and Stefan Teipel

Oil in the Machine: Technical Support for a Human-CentredService System for Public Transport. . . . . . . . . . . . . . . . . . . . . . . . . . 157Jan Alexandersson, David Banz, Daniel Bieber, Jochen Britz,Maurice Rekrut, Kathleen Schwarz, Florin Spanachi,Martin Thoma and Johannes Tröger

x Contents

Part VI Community Conclusions

Quantitative and Qualitative Rating and RankingStudies for Consolidation of an Application Portfoliofor Large Scale Pilots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171Henk Herman Nap, Ad van Berlo and Reiner Wichert

Contents xi

Part IInteraction and Robotics

Challenges in Adopting Speech Controlfor Assistive Robots

Paul Panek and Peter Mayer

Abstract This paper presents a pragmatic report about experiences gathered withthe speech command interface of HOBBIT, a mobile assistive robot intended tosupport older persons in their private home. An outline of the associated problemsand challenges in distant speech recognition is given. Far field detection of user’svoice commands is necessary due to user acceptance considerations. A commerciallyavailable automatic speech recognition (ASR) serves as base. Measurements ofdirectivity of several microphones were done and ASR performance with a smallvocabulary showed low word error rates (WER) in different acoustic environments:anechoic room (0%), free space (2.6 %), AAL room (3.8%) when using a small arraymicrophone with beam forming. Further explorative trials in a free space setting witha second (disturbing) signal source resulted in a WER of 3.9 % (two voices) and11.1 % (one voice and radio news) compared to 2.6 % in case of only one speaker.

Keywords AAL � Socially assistive robots � Speech recognition � HRI

1 Introduction and Aim

The increasing percentage of old persons within the population and the forecastacceleration rate until 2050 [1] gave rise to many publications and projects dealingwith solutions for the associated problems. One of the many scenarios researched intois the application of social assistive robotics in home or institutional settings [2–5].

P. Panek (&) � P. MayerCentre for Applied Assistive Technologies (AAT), Institute for Design and Assessmentof Technology, Vienna University of Technology, Favoritenstrasse 11/187-2b,1040 Vienna, Austriae-mail: panek@fortec.tuwien.ac.at

P. Mayere-mail: mayer@fortec.tuwien.ac.at

© Springer International Publishing Switzerland 2015R. Wichert and H. Klausing (eds.), Ambient Assisted Living,Advanced Technologies and Societal Change,DOI 10.1007/978-3-319-11866-6_1

3

The process of increased automation alone does not automatically guaranteeimproved human-machine system performance. Poor use of technology oftenresults in systems that are difficult to learn or use especially for people who are nocomputer addicts. Therefore, human-computer interaction (HCI) is an establishedresearch topic in itself that becomes human-robot interaction (HRI) for assistiverobots. While the computer in the robot stays just a machine, the expectationstowards these machines, which also are designed to resemble humans and mimictheir behaviour on purpose [4], are even increasing. A key point to mimic human-human communication is the communication with the robot via speech.

The speech output side is nowadays well developed with many Text-To-Speech(TTS) solutions which meanwhile sound rather human. Automatic speech recog-nition (ASR) however, while found working satisfactory in text dictation or time-table enquiry tasks or in other well specified close talk conditions, is still a challengewhen human and robot share a larger operating area in real-life environments.

No literature seems to be available with substantiated data regarding microphoneconfiguration and overall ASR performance data in realistic setting suitable for ourrather pragmatic needs. This motivated us to revisit the known challenges in thearea of distant speech recognition and to initiate some first explorative trials indifferent acoustic settings.

2 The HOBBIT Project

The HOBBIT robot (Fig. 1) provides autonomous navigation, a manipulator with agripper and a multi-modal user interface (UI) allowing interaction via speech,gesture and touch screen [6]. The UI provides easy access to information in the

Fig. 1 HOBBIT prototype 1(with and without cover) [6]

4 P. Panek and P. Mayer

web, videophone service, serious games, control of robot functions (e.g. themanipulator), emergency call features and control of the AAL environment in anaccessible and consistent way [7]. Additionally, a small display on top of the robotpresents emotions by expression of eye and mouth.

The autonomous robot as the mobile element of HOBBIT shares informationwith the intelligent environment in which it is operating in order to establishenriched context awareness. Also, the robot can make use of actuators in theenvironment to extend its capabilities [8].

2.1 Speech Control in HOBBIT

The speech recognition in HOBBIT has to be provided in four languages, three ofwhich (German, Greek, Swedish) are used in user trials and one (English) is used asdevelopment master and for presentations. Because of the need for four languagesafter some initial tests with several Automatic Speech Recognition (ASR) enginesthe decision was made to choose a commercial speaker-independent ASR enginewith no need for collecting and processing speech samples for training in all lan-guages. It is clear that with individual training (also to the special voices of oldpeople) the recognition performance could be further optimized at great expense,but it can be expected that some improvements will also occur because of theadaptive features of the ASR during usage.

Within the project there are no resources for developing NLU (Natural LanguageUnderstanding) in four languages or special soft- and hardware, it is not foreseen toinstall microphones in the rooms, and many other projects have targeted this already[9–11]. A simpler approach based on a commercial ASR engine with good supportfor translation (Loquendo/Nuance) was chosen and successfully applied duringprototype 1 trials [12].

HOBBIT makes use of a context free grammar in W3C SRGS format withsemantic annotation by tags. As part of the multi-modal UI approach every com-mand that is available with caption on the touch screen interface menus can also begiven by speech input 1:1 and in several variations.

The multi-modal input comprises of gestures, speech or touch which are fused inthe first stage of the dialogue manager to solve dependencies, conflicts andambiguities.

2.2 Acoustic Challenges in Distant Speech

Social assistive robots like HOBBIT act autonomously with-in the whole apartmentof the user. Typically, the acoustic situation changes with every different room andfor every position within a room. While for typical dictation tasks the user speaksinto a headset microphone or into a very close desktop microphone for HOBBIT the

Challenges in Adopting Speech Control for Assistive Robots 5

distance to the user might vary from 0.5 m to several meters and the user not alwayswill be oriented towards the robot’s microphone. When the distance betweenspeaker and microphone increases the level of the directly received signal decreasesand effects of reverberation on all objects and walls are becoming more important[13]. The sound waves approaching the receiver via different paths can be char-acterised in three categories:

• direct wave: reaching the receiver on the direct path,• early reflections: reaching the receiver (the ear or the microphone) on an indirect

path and arrive there approximately 50–100 ms after the direct wave,• late reflections: result in a so-called diffuse noise field.

Additionally, other acoustic sources like background noise from the street,household appliances or sound from a switched-on TV set mix up with the user’svoice command. The distortion of desired signal is much more significant in the farfield recording setting, as e.g. the distance from noise source to microphone easilycan be shorter than the distance from the speaker’s mouth to the microphone.

It has to be emphasised that contemporary ASR systems work well for differentapplications, but only as long as the microphone is not moved away from thespeaker’s mouth. The latter case is called distant or far-field speech recognition andshows a significant drop in performance [14–16], which mainly is due to threedifferent types of distortion [13]:

• background noise (any sound other than the desired speech)• echo and reverberation (reflections of the sound source arriving some time after

the signal on the direct path)• other types of distortions (e.g. environmental factors such as room modes, the

orientation of the speaker’s head, or the Lombard effect)

Human hearing is rather good in selecting only a single source out of suchmixture (party effect) but this is not the case for typical ASR engines. A typicalASR engine is either trained to a specific user (speaker dependent), or on a large setof samples from many speakers (speaker independent) but nearly always thesamples are recorded under very clear conditions with good SNR (signal to noiseratio). The comparison with the actual input from the microphone during the rec-ognition task therefore cannot cope well with the manifold variations to the speechcharacteristics that might occur in real use under non-ideal varying conditions.

In our project, another difference to typical ASR applications is the often lowerand less clear and repeatable voice of old persons further lowering the SN ratio andrecognition rate.

2.3 Acoustic Countermeasures

The first stages of the ASR already cover the adaptation of the input volume withvarying distance. This however, does at the same time apply to all unwanted

6 P. Panek and P. Mayer

signals. It is therefore important to think about improvements to the captured signalbefore it enters the ASR.

2.3.1 Microphone

For sure the microphone is the key element in speech recognition. Most companiesselling ASR engines for dictation purposes also give recommendations for approvedmicrophones (usually headset). As the comparison of an utterance to the storedsamples in the ASR is made after extraction of characteristic speech properties [17] itis above all important to avoid any distortions of the captured wanted signal over thewhole frequency range, be it from noise, reverberation, echo or processing artefactsor amplification, which most easily is achieved by close distance between mouth andmicrophone. The specific construction of a microphone can however modify thedirectional sensitivity characteristics (beam forming). A shotgun microphone forexample by construction is more sensitive to sound from the front than from the sideor rear. This helps to reduce the amount of (mostly unwanted) signals received fromdirections other than the one where the microphone is headed to.

2.3.2 Beam Forming

Nowadays so-called beam forming is also possible by computational pre-processingof the signal from several microphones. If two or more microphones (microphonearray) separated by a certain distance receive the same signal, the small differencesin time of arrival of the signals (because of the different paths the acoustic waves aretaking), can be used to exclude parts of the signal that stem from outside a certainreception angle allowing to adjust the resulting beam width of a virtual microphone.Again the limited reception angle reduces unwanted signals but also requires moredirected alignment of microphone and source. Beam forming works the better themore microphones are used and the wider an area the microphones of the array arespanning. The minimum number of microphones for this purpose is 2 and theminimum distance between microphones for linear arrays is about 10–20 cm whichsets the overall size. Other solutions like the Kinect for the same purpose use 4microphones in a linear array giving more possibility to estimate the wanted signalout of the distorted signal. Good multi-microphone arrays for multi-person con-ferencing applications usually have a minimum length of 50–100 cm or arearranged in a circular or spherical pattern. Adaptive beam forming with manymicrophones even allows focusing dynamically depending on the estimation ofunwanted noise and main wanted source properties.

Conclusions that can be drawn from literature on beam forming:

• Beam forming with microphone arrays is able to enhance speech captured fromdistance but limits the reception angle. Performance can approach that fromclose talk conditions

Challenges in Adopting Speech Control for Assistive Robots 7