Abstract: A method segments a compressed video by extracting audio and visual features from the compressed video. The audio features are clustered according to K-means clustering in a set of classes, and the compressed video is then partitioned into first segments according to the set of classes. The visual features are then used to partitioning each first segment into second segments using motion analysis. Summaries of the second segments can be provided to assist in the browsing of the compressed video.

Abstract: A method detects objects in a scene over time. Sets of time-aligned features are extracted from multiple signals representing a scene over time; each signal is acquired using a different modality. Each set of time-aligned features is arranged as a vector in a matrix to which a first transform is applied to produce a compressed matrix. A second transform is applied to the compressed matrix to extract spatio-temporal profiles of objects occurring in the scene.

Abstract: A method selects recordings stored in a database. A spoken query is represented as a phonetic lattice and paths through the phonetic lattice are converted to a set of text queries. The database is searched to generate a playlist of recordings matching the set of text queries and samples of the recordings on the playlist are then played. A particular sample is selected as an acoustic query for searching the database to update the playlist with recording matching the acoustic query. Samples of the recordings on the updated playlist are played and a particular sample of the updated play list is selected. A particular record associated with the sample is then played.

Abstract: A method segments a compressed video by extracting audio and visual features from the compressed video. The audio features are clustered according to K-means clustering in a set of classes, and the compressed video is then partitioned into first segments according to the set of classes. The visual features are then used to partitioning each first segment into second segments using motion analysis. Summaries of the second segments can be provided to assist in the browsing of the compressed video.

Abstract: A computerized method extracts features from an acoustic signal generated from one or more sources. The acoustic signal are first windowed and filtered to produce a spectral envelope for each source. The dimensionality of the spectral envelope is then reduced to produce a set of features for the acoustic signal. The features in the set are clustered to produce a group of features for each of the sources. The features in each group include spectral features and corresponding temporal features characterizing each source. Each group of features is a quantitative descriptor that is also associated with a qualitative descriptor. Hidden Markov models are trained with sets of known features and stored in a database. The database can then be indexed by sets of unknown features to select or recognize like acoustic signals.

Abstract: A method extracts features from a mixture of signals. The method filters the mixture of signals by a filterbank to produce a plurality of band-pass signals. Each band-pass signal is windowed to produce a plurality of multi-dimensional observation matrices. The multi-dimensional observation matrices are reduced in their dimensionality. Features are extracted from the reduced dimensionality matrices using independent component analysis. The features can include temporal and spectral characteristics.