Over 90% of information comes from the human visual channel. We develop high performance visual computing algorithms and techniques that analyze visual content, recognize human characteristics, visualize pre-processed model and real-world content, and enable users to interact computer generated visual content as well as the real 3D physical world as a single unified environment. We develop parallel algorithms with multi-core and many core machines for efficient 3D visual information processing, such as 3D graphics, 3D sensing and interaction, and video processing. The developed 3D artMap technology is the world 1st 3D artistic navigation map.

Natural language is the primary means for humans to communicate. It enables highly efficient, convenient and safe user interfaces and mutual understanding. We develop and apply advanced machine learning and semantic modeling methods in areas such as spoken and text language understanding, dialog management, response generation, knowledge management, information extraction, and document content analysis. We prototype superior speech, multimodal, and intelligent systems for domains important to Bosch businesses.

Developing a superior user interaction system is costly and takes long time, as new applications typically have new requirements. In this line of research, we explore ways to efficiently and cost-effectively adapt latest user interaction technologies into different software and hardware platforms. We also develop technologies to ensure transient user experience across different platforms from server-based cloud, to mobile computing, and to embedded devices, based on the latest web technologies, such as HTML 5.

We develop efficient machine learning and data mining algorithms and techniques for large scale processing and modeling of human explicit and implicit communication data in areas, such as user modeling, reasoning, behavior prediction, and personalization.