Robust Speech Processing (RoSP)

Click here to visit the SIG website! 

Aims. RoSP aims at providing the following services:

• to promote interest in Robust Speech Processing;
• to provide members of ISCA with a special interest in Robust Speech Processing with a means of exchanging news of recent research developments and other matters of interest in Robust Speech Processing;
• to promote better understanding of the robustness challenges posed by real-world speech applications at various levels, including (but not limited to) the acoustic level (channel change, reverberation, background noise, overlapping speech...), the speaker level (stressed speech, Lombard speech, speaker change, change of speaking rate, accent...) and the language level (dialect, topic change...);
• to facilitate the development of theoretical foundations and practical techniques addressing these challenges by means of shared goals and terminology;
• to call attention to the applications of these techniques, including (but not limited to) automatic speech recognition, diarization, speaker identification and verification, spoken document retrieval, dialogue and spoken language understanding, speech-to-speech translation, speaker localization, speech separation and enhancement, human-machine interaction and robust training for speech synthesis;
• to create a scientific link between researchers and others working in the various disciplines associated with Robust Speech Processing;
• to provide and make available resources relevant to Robust Speech Processing, including speech and text corpora, best-practice processing and evaluation methodologies, processing and evaluation software, research papers and generated data;
• to encourage diversity within the speech processing field and promote inclusivity by making sure that researchers from different backgrounds, cultures and genders feel welcomed and have equal opportunities to contribute;
• to propose reviewers for ISCA conferences and workshops and to nominate outstanding members for ISCA awards;
• and to sponsor meetings, workshops and evaluation campaigns in Robust Speech Processing that appear to be timely and worthwhile, operating within the framework of ISCA's by-laws for SIGs.

Motivation. Along with the rest of the speech processing community, robust speech processing has experimented strong changes in the last years. These changes have been driven by the maturity of fields like automatic speech recognition (ASR) which has been labeled an strategic asset by some of the biggest and more research-active players in the new technologies market. Speech-enabled applications are now more pervasive than ever, with the rise of speech-enabled virtual assistants, meeting transcription, generative AI and hands-free telecommunication to name a few, and, as such, robust speech processing techniques are also increasingly important. This growth calls for new organizational forms that can help the coordination within the community and allow it to adapt to this new dynamic scenario. The main goal of RoSP-SIG is to strengthen the ties between the different research sub-fields involving robust speech processing and coordinate actions in the interest of the whole community such as increasing resource availability, the creation of guidelines and the organization of specific workshops and challenges. Going beyond, we also highly encourage interdisciplinary connections that can foster advance of the field.

• Chair: Paola Garcia  (Johns Hopkins University, USA)
• ISCA Liaison Officer: Samuele Cornell (Carnegie Mellon University, USA)

ISCA-supported events. RoSP supports the series of workshops on Hands-free Speech Communication and Microphone Arrays (HSCMA):

• HSCMA 2024, Seoul, South Korea
• HSCMA 2017, San Francisco, USA
• HSCMA 2014, Nancy, France
• HSCMA 2011, 30/5-1/6, Edinburgh
• HSCMA 2008, 6/5-8/5, Trento, Italy
• HSCMA 2005, Rutgers University, NJ

RoSP supports also other events at major speech conferences included the recent CHiME (Computational Hearing in Multisource Environments) Challenges and Workshops:

• CHiME Speech Separation and Recognition Challenges and International Workshops (2015-present)
• Robust Speech Processing using Observation Uncertainty and Uncertainty Propagation, at Interspeech 2015, Dresden, Germa

To join, just register to our Google group

You are also welcome to join our LinkedIn group