BioS-DB: a multimodal database of individuals in a public speaking scenario, including emotional annotation

<p>The BioS-DB (BioSpeech Database), is a database of indivduals speaking infront of others in both German and English. BioS-DB includes 55 indivdual (33 male and 22 female), with a mean age of 28.9 years ( &plusmn; 10.5 years). Individuals were predominately German Natives (33) - and either students (30) or staff from the computer science department at the University of Augsburg, Germany. The average speech length was 45 s for German and 42 s for English.</p> <p>During the speech, individuals were being evaluated for their emotion (valence / arousal) in a time-continuous way. Individuals were also attached to Blood Volume Pulse, and Skin Conductance sensors, while audio was captured from a lapel microphone and additionally a room microphone.</p> <p>Alice Baird, Shahin Amiriparian, Miriam Berschnider, Maximilian Schmitt, and Bj&ouml;rn Schuller (2019),<em> Predicting Biological Signals from Speech: Introducing a Novel Multimodal Dataset and Results</em>, The Multimodal Signal Processing Conference, Kuala Lumpur, Malaysia, Sept 2019. 5 pages.</p> <p><strong>Version 2.0:&nbsp;</strong>This version contains the raw biological signal, and the individual&nbsp;annotations for re-computing the gold-standard.&nbsp;</p> <p>Alice Baird, Shahin Amiriparian, Manuel Milling,&nbsp;Bj&ouml;rn Schuller (2020),&nbsp;<em>Emotion Recognition in Public Speaking Scenarios Utilising an LSTM-RNN Approach With Attention,&nbsp;</em>The Speech Language Technology Conference, held virtually (to appear) Jan. 2021. 5 pages.&nbsp;</p>