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:20251015T222002Z UID:0A7214D4-F700-4914-9DE4-C8D080FFBF72 DTSTART;TZID=America/Chicago:20251015T100000 DTEND;TZID=America/Chicago:20251015T110000 DESCRIPTION:Title: Convolutional Beamformer for Joint Denoising\, Dereverbe ration\, and Source Separation\n\nAbstract:\nWhen speech is captured by di stant microphones in everyday environments\, the signals are often contami nated by background noise\, reverberation\, and overlapping voices. The co nvolutional beamformer (CBF) is a signal processing technique that recover s clean\, close-microphone-quality speech from such complex mixtures. By j ointly performing denoising\, dereverberation\, and source separation\, CB F enhances both human listening experiences and automatic speech recogniti on (ASR) accuracy. Potential applications include hearing assistive device s\, meeting transcription systems\, and other real-world speech technologi es.\n\nThis talk begins by introducing the concept of CBF\, including its formal definition\, mechanism for joint enhancement\, and optimization via maximum likelihood estimation. CBF is defined as a series of beamformers estimated at each frequency in the short-time Fourier transform (STFT) dom ain and convolved with the observed signal to achieve the desired enhancem ent. The presentation then describes that CBF can be factorized into Multi channel Linear Prediction (MCLP) for dereverberation and Beamforming (BF) for denoising and separation\, highlighting the practical advantages of th is decomposition. Related work is reviewed\, including Weighted 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.\n\nFurther extensions are presented\, i ncluding blind CBF for unknown recording conditions\, switching CBF for en hanced performance with a limited number of microphones\, and integration with neural networks - notably the DiffCBF framework\, which combines CBF with diffusion-based speech enhancement models. Experimental results demon strate state-of-the-art speech quality\, even with relatively few micropho nes and limited training data.\n\nCo-sponsored by: Starkey\n\nSpeaker(s): Dr. Tomohiro Nakatani\n\nAgenda: \n9:30 – 10:00 a.m. Meet and Greet\n\n1 0:00 – 10:05 a.m. Welcome Remarks by Dr. Masahiro Sunohara\n\n10:05 – 11:00 a.m. Convolutional Beamformer for Joint Denoising\, Dereverberation\ , and Source Separation by Dr. Tomohiro Nakatani\n\nRoom: Excelsior Room\, Bldg: William F Austin Center\, 6425 Flying Cloud Dr\, Eden Prairie\, Min nesota\, United States\, 55344\, Virtual: https://events.vtools.ieee.org/m /501955 LOCATION:Room: Excelsior Room\, Bldg: William F Austin Center\, 6425 Flying Cloud Dr\, Eden Prairie\, Minnesota\, United States\, 55344\, Virtual: ht tps://events.vtools.ieee.org/m/501955 ORGANIZER:masahiro_sunohara@starkey.com SEQUENCE:64 SUMMARY:IEEE SPS DISTINGUISHED INDUSTRY SPEAKER PROGRAM TWIN CITIES SP/COM CHAPTER SEMINAR 10/15/2025 URL;VALUE=URI:https://events.vtools.ieee.org/m/501955 X-ALT-DESC:Description: <br /><p>Title:&nbsp\;<span style="text-decoration: underline\;"><strong>Convolutional Beamformer for Joint Denoising\, Derev erberation\, and Source Separation</strong></span></p>\n<p>Abstract:<br>Wh en speech is captured by distant microphones in everyday environments\, th e signals are often contaminated by background noise\, reverberation\, and overlapping voices. The convolutional beamformer (CBF) is a signal proces sing technique that recovers clean\, close-microphone-quality speech from such complex mixtures. By jointly performing denoising\, dereverberation\, and source separation\, CBF enhances both human listening experiences and automatic speech recognition (ASR) accuracy. Potential applications inclu de hearing assistive devices\, meeting transcription systems\, and other r eal-world speech technologies.</p>\n<p>This talk begins by introducing the concept of CBF\, including its formal definition\, mechanism for joint en hancement\, and optimization via maximum likelihood estimation. CBF is def ined as a series of beamformers estimated at each frequency in the short-t ime Fourier transform (STFT) domain and convolved with the observed signal to achieve the desired enhancement. The presentation then describes that CBF can be factorized into Multichannel Linear Prediction (MCLP) for derev erberation and Beamforming (BF) for denoising and separation\, highlightin g the practical advantages of this decomposition. Related work is reviewed \, including Weighted Prediction Error (WPE) dereverberation\, mask-based beamforming\, and guided source separation\, with emphasis on strong resul ts in challenging tasks such as the CHiME-8 distant ASR challenge.</p>\n<p >Further extensions are presented\, including blind CBF for unknown record ing conditions\, switching CBF for enhanced performance with a limited num ber of microphones\, and integration with neural networks - notably the Di ffCBF framework\, which combines CBF with diffusion-based speech enhanceme nt models. Experimental results demonstrate state-of-the-art speech qualit y\, even with relatively few microphones and limited training data.</p><br /><br />Agenda: <br /><p>9:30 &ndash\; 10:00 a.m.&nbsp\; &nbsp\; &nbsp\; &nbsp\;<strong>Meet and Greet&nbsp\;</strong></p>\n<p>10:00 &ndash\; 10:05 a.m.&nbsp\; &nbsp\; &nbsp\;<strong>Welcome Remarks</strong>&nbsp\;by&nbsp \;<em>Dr. Masahiro Sunohara</em></p>\n<p>10:05 &ndash\; 11:00 a.m.&nbsp\; &nbsp\; &nbsp\;<strong>Convolutional Beamformer for Joint Denoising\, Dere verberation\, and Source Separation</strong> by Dr. Tomohiro Nakatani</p> END:VEVENT END:VCALENDAR