University of Twente
Partner Contributions
- Fusion of speech and textual content English, Dutch and German
- Word spotting in order to recognize brands and company names with high probability
- Jingle recognition (especially for advertisement campaigns jingles are related to certain companies)
The multi-disciplinary research on information and communication technology of the University of Twente is concentrated in the Centre of Telematics and Information Technology (CTIT). As such it combines knowledge of ICT systems and applications. The Centre has liaisons with large international ICT companies and European research institutes. The Human Media Interaction (HMI) research team of the CTIT is responsible for the research areas multimodal interaction, content-based multimedia access and multimedia presentation, ambient intelligence, intelligent and social interfaces, and machine learning. Within HMI, research is currently being carried out in the following areas which range from truly fundamentl to highly application oriented: sensing and interpreting (multi-party) human verbal and non-verbal communication behaviour, annotation tools, models for information fusion, speech and language processing, emotion detection (voice, facial expression), affect in interaction, multimedia indexing, searching and summarization, embodied conversational agents, animation, virtual reality environments for experiments, and education, etc.
HMI participated in various EU research projects on multimedia retrieval such as the 4th framework projects OLIVE, POP-EYE and TWENTYONE and the IST projects ECHO (European CHronicles Online) and MUMIS (Multimedia Indexing and Searching Environment). In the 5th HMI was involved in M4, a project on capturing and disclosing A/V recordings from sessions in a smart meeting room. In the 6th FP HMI takes part in IP AMI on capturing and interpreting human multi-party interaction in a smart meeting room. In the national context, the group participates in funded projects on information presentation by embodied conversational agents, speech-supported video retrieval, dialogue systems, information extraction and multimodal interaction for collaborative work and decision-support environments. Within these projects HMI has worked together with both technology providers, as well as content owners, such as broadcast archives and news publishers. Whenever appropriate, performance evaluation is done via of participation in international benchmarks for information processing tasks such as TREC, CLEF, TRECVID, TDT.
HMI is/was involved in a number of networks of excellence sponsored by the EU: DELOS (Digital Libraries), ELSNET (language and speech processing), INTUITION (virtual reality), HUMAINE (Human-Machine Interaction Network on Emotion) and SCHEMA.