BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Chicago BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0600 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:CDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20251102T010000 TZOFFSETFROM:-0500 TZOFFSETTO:-0600 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20251202T010041Z UID:66A3C196-909D-45C4-A29F-DDEAAD9AA197 DTSTART;TZID=America/Chicago:20251017T183000 DTEND;TZID=America/Chicago:20251017T200000 DESCRIPTION:When speech is captured by distant microphones in everyday envi ronments\, the signals are often contaminated by background noise\, reverb eration\, and overlapping voices. The convolutional beamformer (CBF) is a signal processing technique that recovers clean\, close-microphone-quality speech from such complex mixtures. By jointly performing denoising\, dere verberation\, and source separation\, CBF enhances both human listening ex periences and automatic speech recognition (ASR) accuracy. Potential appli cations include hearing assistive devices\, meeting transcription systems\ , and other real-world speech technologies. This talk begins by introducin g the concept of CBF\, including its formal definition\, mechanism for joi nt enhancement\, and optimization via maximum likelihood estimation. CBF i s defined as a series of beamformers estimated at each frequency in the sh ort-time Fourier transform (STFT) domain and convolved with the observed s ignal to achieve the desired enhancement. The presentation then describes that CBF can be factorized into Multichannel Linear Prediction (MCLP) for dereverberation and Beamforming (BF) for denoising and separation\, highli ghting the practical advantages of this decomposition. Related work is rev iewed\, including Weighted Prediction Error (WPE) dereverberation\, mask-b ased beamforming\, and guided source separation\, with emphasis on strong results in challenging tasks such as the CHiME-8 distant ASR challenge. Fu rther extensions are presented\, including blind CBF for unknown recording conditions\, switching CBF for enhanced performance with a limited number of microphones\, and integration with neural networks - notably the DiffC BF framework\, which combines CBF with diffusion-based speech enhancement models. Experimental results demonstrate state-of-the-art speech quality\, even with relatively few microphones and limited training data.\n\nSpeake r(s): Tomohiro Nakatani\, Ph.D.\n\nAgenda: \n6:30 - 7:00 Social half hour to grab food and drink\n\n7:00 - 8:00 Technical talk\n\nRoom: Mann Hall\, Bldg: Medical Sciences Building\, 300 3rd Ave SW\, Rochester\, Minnesota\, United States\, 55902\, Virtual: https://events.vtools.ieee.org/m/499193 LOCATION:Room: Mann Hall\, Bldg: Medical Sciences Building\, 300 3rd Ave SW \, Rochester\, Minnesota\, United States\, 55902\, Virtual: https://events .vtools.ieee.org/m/499193 ORGANIZER:pramanik.leena@ieee.org SEQUENCE:15 SUMMARY:October DISTINGUISHED INDUSTRY SPEAKER Talk: Convolutional Beamform er for Joint Denoising\, Dereverberation\, and Source Separation (HYBRID) URL;VALUE=URI:https://events.vtools.ieee.org/m/499193 X-ALT-DESC:Description: <br /><p style="text-align: justify\;">When speech is captured by distant microphones in everyday environments\, the signals are often contaminated by background noise\, reverberation\, and overlappi ng voices. The convolutional beamformer (CBF) is a signal processing techn ique that recovers clean\, close-microphone-quality speech from such compl ex mixtures. By jointly performing denoising\, dereverberation\, and sourc e separation\, CBF enhances both human listening experiences and automatic speech recognition (ASR) accuracy. Potential applications include hearing assistive devices\, meeting transcription systems\, and other real-world speech technologies. This talk begins by introducing the concept of CBF\, including its formal definition\, mechanism for joint enhancement\, and op timization via maximum likelihood estimation. CBF is defined as a series o f beamformers estimated at each frequency in the short-time Fourier transf orm (STFT) domain and convolved with the observed signal to achieve the de sired enhancement. The presentation then describes that CBF can be factori zed into Multichannel Linear Prediction (MCLP) for dereverberation and Bea mforming (BF) for denoising and separation\, highlighting the practical ad vantages of this decomposition. Related work is reviewed\, including Weigh ted Prediction Error (WPE) dereverberation\, mask-based beamforming\, and guided source separation\, with emphasis on strong results in challenging tasks such as the CHiME-8 distant ASR challenge. Further extensions are pr esented\, including blind CBF for unknown recording conditions\, switching CBF for enhanced performance with a limited number of microphones\, and i ntegration with neural networks - notably the DiffCBF framework\, which co mbines CBF with diffusion-based speech enhancement models. Experimental re sults demonstrate state-of-the-art speech quality\, even with relatively f ew microphones and limited training data.</p><br /><br />Agenda: <br /><p> <em>6:30 - 7:00</em>&nbsp\;Social half hour to grab food and drink</p>\n<p ><em>7:00 - 8:00</em>&nbsp\;Technical talk</p>\n<p>&nbsp\;</p> END:VEVENT END:VCALENDAR