Output details
11 - Computer Science and Informatics
Imperial College London
Being bored? Recognising natural interest by extensive audiovisual integration for real-life application
<23>For the first time, acoustic and linguistic cues together with non-linguistic cues such as laughter, and vision cues were recognized and fused fully automatically to assess human interest. From a Machine Learning perspective, fusion by functional application across the diverse modalities is introduced as highly efficient solution to cope with asynchrony. The software is exploited by Toyota Inc. and a demonstrator runs in a robot museum guide in Japan that recognises visitors’ (dis-)interest to change the exhibit. The collected data and principle was later used for the highly successful INTERSPEECH 2010 Paralinguistic Challenge where we still prevail.