BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Europe/Copenhagen
BEGIN:DAYLIGHT
DTSTART:20260329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260222T063151Z
UID:B470FBBD-04FD-42FC-BF24-3B7747A13552
DTSTART;TZID=Europe/Copenhagen:20260219T090000
DTEND;TZID=Europe/Copenhagen:20260219T110100
DESCRIPTION:&quot;Multimodal Unsupervised Speech Enhancement via Probabilistic S
 peech Priors&quot;\n\nby\nXavier Alameda-Pineda\n\nInria\, University Grenoble 
 Alpes\, France\n\nAbstract: In this talk we will explore different probabi
 listic speech priors that exploit audio-visual data\, for the task of unsu
 pervised speech enhancement. In short\, unsupervised speech enhancement is
  the task of improving the quality of the speech signal without access to 
 noise samples during training. Therefore\, training is done exclusively wi
 th clean speech signals. At test time\, the double task of estimating the 
 noise parameters and the clean speech signal from a noisy observation is a
 ddressed. We will discuss the role of the visual modality in building a sp
 eech prior that is more robust to noise\, allowing for better noise parame
 ter estimates\, and overall improved speech enhancement performance.\n\nan
 d\n\n“Recent Advances in Speech Deepfakes: From Detection to Source Trac
 ing and Spoofing-Robust Speaker Verification”\n\nby\nTomi Kinnunen\n\nUn
 iversity of Eastern Finland\, Finland\n\nAbstract: We have seen a surge in
  voice cloning services that anyone can use nowadays to craft synthetic vo
 ices. Motivated by security considerations of calls\, teleconfercing\, aud
 io in social media and protecting integrity of voice biometric (automatic 
 speaker verification)\, the research community has worked over a decade fo
 r novel solutions for detecting deepfakes. Recently\, there is also an inc
 reasing interest in determining the origin of a deepfake (e.g. a particula
 r speech synthesis system). In this talk\, I provide a selective summary o
 f the background in this emerging field\, along with brief summary of rece
 nt advances\, including findings from the ASVspoof 5 challenge edition.\n\
 nThursday February 19\, 2026\, at 09:00\n\nAalborg University\, Fredrik Ba
 jers Vej 7A4-108\n\nRoom: A4-108\, Fredrik Bajers Vej 7\, Aalborg\, Nordjy
 llands Amt\, Denmark
LOCATION:Room: A4-108\, Fredrik Bajers Vej 7\, Aalborg\, Nordjyllands Amt\,
  Denmark
ORGANIZER:zt@es.aau.dk
SEQUENCE:17
SUMMARY:Invited talks on multimodal unsupervised speech enhancement and spe
 ech deepfakes
URL;VALUE=URI:https://events.vtools.ieee.org/m/540181
X-ALT-DESC:Description: &lt;br /&gt;&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;strong&gt;&lt;
 span lang=&quot;EN-US&quot;&gt;&quot;Multimodal Unsupervised Speech Enhancement via Probabil
 istic Speech Priors&quot;&lt;/span&gt;&lt;/strong&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;&amp;nbsp\;&lt;/sp
 an&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;span lang=&quot;EN-US&quot;&gt;b
 y&lt;br&gt;&lt;/span&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;&lt;span class=&quot;searchHighlight&quot;&gt;Xavie
 r&lt;/span&gt; Alameda-Pineda&lt;/span&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;c
 enter&quot;&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;Inria\, University Grenoble Alpes\, Fran
 ce&lt;/span&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p align=&quot;center&quot;&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;&lt;stro
 ng&gt;Abstract: &lt;/strong&gt;&lt;/span&gt;&lt;/strong&gt;In this talk we will explore differe
 nt probabilistic speech priors that exploit audio-visual data\, for the ta
 sk of unsupervised speech enhancement. In short\, unsupervised speech enha
 ncement is the task of improving the quality of the speech signal without 
 access to noise samples during training. Therefore\, training is done excl
 usively with clean speech signals. At test time\, the double task of estim
 ating the noise parameters and the clean speech signal from a noisy observ
 ation is addressed. We will discuss the role of the visual modality in bui
 lding a speech prior that is more robust to noise\, allowing for better no
 ise parameter estimates\, and overall improved speech enhancement performa
 nce.&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;span lang=&quot;EN-US&quot;&gt;and&lt;/span
 &gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;&amp;ldq
 uo\;&lt;/span&gt;&lt;/strong&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;Recent Advances in Speech D
 eepfakes: From Detection to Source Tracing and Spoofing-Robust Speaker Ver
 ification&amp;rdquo\;&lt;/span&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;
 &gt;&lt;span lang=&quot;EN-US&quot;&gt;by&lt;br&gt;&lt;/span&gt;&lt;strong&gt;&lt;span lang=&quot;EN-US&quot;&gt;Tomi Kinnunen&lt;
 /span&gt;&lt;/strong&gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;strong&gt;&lt;span lan
 g=&quot;EN-US&quot;&gt;University of Eastern Finland\, Finland&lt;br&gt;&lt;/span&gt;&lt;/strong&gt;&lt;/p&gt;\
 n&lt;p class=&quot;MsoNormal&quot; align=&quot;center&quot;&gt;&lt;strong&gt;Abstract: &lt;/strong&gt;We have se
 en a surge in voice cloning services that anyone can use nowadays to craft
  synthetic voices. Motivated by security considerations of calls\, telecon
 fercing\, audio in social media and protecting integrity of voice biometri
 c (automatic speaker verification)\, the research community has worked ove
 r a decade for novel solutions for detecting deepfakes. Recently\, there i
 s also an increasing interest in determining the origin of a deepfake (e.g
 . a particular speech synthesis system). In this talk\, I provide a select
 ive summary of the background in this emerging field\, along with brief su
 mmary of recent advances\, including findings from the ASVspoof 5 challeng
 e edition.&lt;/p&gt;\n&lt;p class=&quot;MsoNoSpacing&quot; align=&quot;center&quot;&gt;&lt;span lang=&quot;EN-US&quot;&gt;
 Thursday February 19\, 2026\, at 09:00&lt;/span&gt;&lt;/p&gt;\n&lt;p class=&quot;MsoNoSpacing&quot;
  align=&quot;center&quot;&gt;Aalborg University\, Fredrik Bajers Vej 7A4-108&lt;/p&gt;
END:VEVENT
END:VCALENDAR