Multimedia Pattern Recognition

Here, you find the home page of the Multimedia Pattern Recognition research group of Fraunhofer IAIS. Our research group investigates and develops advanced methods and algorithms for the segmentation, recognition, and indexing of speech, audio, image, and video data. The group is headed by Dr. Joachim Köhler who is also head of the department NetMedia.

The motivation of our research work is to investigate new methods in the area of signal processing and pattern recognition for multimedia data sources to enable the automatic analysis of growing content repositories.

Multimedia pattern recognition applies data-depending signal analysis methods to extract relevant features for pattern recognition and classification. Speech processing, for instance, develops methods for recognizing speech and to identify different speakers. Image and video analysis develop algorithms for object recognition and segmentation. These algorithms can also be used to analyze image data in print documents. Highlights of our recent research work are:

• syllable based subword speech indexing and retrieval
• automatic segmentation of scanned documents
• face detection and indexing

Selected application scenarios of our research work are multimedia search and indexing and human computer interaction systems.