Program

  - Tuesday, July 17:    08:50-18:40 technical session,
  - Wednesday, July 18:  08:50-12:45 technical session,
                         14:00-21:00 social event (*),
  - Thursday, July 19:   08:50-18:00 technical session.
  (*) Please note that the social event is a boat trip, so you cannot leave in between.

Tuesday, July 17

08:50	Welcome

09:00	Keynote 1

	Chin-Hui Lee, GeorgiaTech, Atlanta, USA:	A machine learning approach for enhancement and recognition of microphone array speech. [Slides]

10:15	Distant Speech Recognition 1

	10:15	Nikos Stefanakis, FORTH, Heraklion, Greece:	Acoustic beamforming in front of a reflective plane. [Slides]
	10:35	Maurizio Omologo, FBK, Trento, Italy:	Realistic impulse responses for distant-speech recognition. [Slides]
	10:55	Patrick Naylor, Imperial College London, UK, and Marcel Katz, Nuance, Germany:	Robust statistical processing of TDOA estimates for distant speaker diarization. [Slides]

11:15	Break

11:45	Keynote 2

	Jon Barker, U Sheffield, UK:	The CHiME-5 challenge for far-field conversational speech recognition in domestic environments. [Slides]

13:00	Lunch

14:15	Keynote 3

	Reinhold Häb-Umbach, U Paderborn, Germany:	Neural network supported acoustic beamforming and source separation for ASR. [Slides]

15:30	End-to-End ASR

	15:30	Albert Zeyer and Ralf Schlüter, RWTH Aachen, Germany:	Attention-based ASR utilizing byte-pair encoding. [Slides]
	15:50	Florian Metze, CMU, Pittsburgh, USA:	Grounded sequence to sequence transduction (multi-modal speech recognition). [Slides]

16:10	Break

16:40	Speaker Separation and Localization

	16:40	Sunit Sivasankaran, INRIA, Nancy, France:	Keyword based speaker localization: Localizing a single speaker in multi-speaker environment. [Slides]
	17:00	Mehdi Zohourian, U Bochum, Germany:	Source localization and separation for binaural hearing aids. [Slides]

17:20	Distant Speech Recognition 2

	17:20	Wei Zhou, EML, Heidelberg, Germany:	The ASR system for the EML LISTEN demonstrator. [Slides]
	17:40	Jozef Ivanécky, EML, Heidelberg, Germany:	Architecture of the EML LISTEN demonstrator. [Slides]
	18:00	Gerasimos Potamianos, U of Thessaly, Greece:	Far-field ASR in Greek for domestic environment and child-robot-interaction applications. [Slides]
	18:20	Pranay Dighe, IDIAP, Martigny, Switzerland:	Improving far-field ASR using low-rank and sparse models. [Slides]

18:40	End of Session

Wednesday, July 18

08:50	Announcements

09:00	Keynote 4

	Jasha Droppo, Microsoft, Redmond, USA:	The Microsoft 2017 conversational speech recognition system. [Slides]

10:15	Current Topics in ASR 1

	10:15	Tim Fingscheidt, TU Braunschweig, Germany:	Acoustic model fusion for phoneme recognition according to the turbo principle. [Slides]
	10:35	Mathew Magimai-Doss, IDIAP, Martigny, Switzerland:	Learning to unlearn and relearn speech signal processing using neural networks: current and future perspectives. [Slides, updated July 19, 2018, 15:22h]
	10:55	Markus Müller, KIT, Karlsruhe, Germany:	Neural modulation for multilingual speech recognition. [Slides]

11:15	Break

11:30	Keynote 5

	Satoshi Nakamura, NAIST, Nara, Japan:	Toward Machine Speech Chain with Semi-supervised Learning by ASR-TTS coupling and Next Generation Speech-to-speech Translation. [Slides]

12:45	Lunch

14:00	Departure Bus Transfer to Social Event

Thursday, July 19

08:50	Announcements

09:00	Keynote 6

	Erik McDermott, Google, Mountain View, USA:	(Towards) next generation acoustic models for automatic speech recognition. [Slides, updated July 24, 2018, 13:28h]

10:15	Distant Speech Recognition 3

	10:15	Igor Szoke, U Brno, Czech Republic:	Collection of re-transmitted data and impulse responses and remote ASR and speaker verification. [Slides]
	10:35	Lukas Drude, U Paderborn, Germany:	Integrating neural network supported beamforming and dereverberation for distant speech recognition. [Slides]
	10:55	Tobias Menne, RWTH Aachen, Germany:	Speaker-adaptive beamforming for ASR. [Slides]

11:15	Break

11:45	Keynote 7

	Alex Acero, Apple, Cupertino, USA:	The Deep Learning Revolution. [Slides (115MB)]

13:00	Lunch

14:15	Keynote 8

	Björn Hoffmeister, Amazon, Seattle, USA:	Building far-field speech recognition for Amazon Alexa: Challenges and solutions. [Slides]

15:30	Current Topics in ASR 2

	15:30	Adrián Giménez and Joan Albert Silvestre, UPV, Valencia, Spain:	Speaker-adapted confidence measures for ASR using deep bidirectional recurrent neural networks. [Slides]
	15:50	Alfons Juan, UPV, Valencia, Spain:	Multilingual videos for education. [Slides]

16:10	Break

16:40	Speaker Separation, Localization and Diarization

	16:40	Dorothea Kolossa, U Bochum, Germany:	Exploiting structures of temporal causality for robust speaker localization in reverberant environments. [Slides]
	17:00	Lauréline Pérotin, INRIA, Nancy, France:	Multichannel RNN-based separation of overlapping speech. [Slides]
	17:20	Katerina Zmolikova, U Brno, Czech Republic:	Speaker aware neural network for speaker extraction from overlapping speech. [Slides]
	17:40	Sanjeev Khudanpur, JHU, Baltimore, USA:	Experiences and lessons learned from the inaugural DIHARD challenge. [Slides]

18:00	End of Session

Zip-archive with all presentation slides (200MB, updated July 24, 2018, 13:28h)

Contact

For organizational issues, especially in case of questions w.r.t. travel and accommodation/hotel booking, please contact our secretariat:
Anna Eva Andersen, Stephanie Jansen, phone: +49 (241) 80-21601 or -21606

During the workshop please call Volker Steinbiss, mobile: +49 (I79) 6679I8