Abstract: A processing system may receive an example image for use in querying a collection of digital images. The processing system may use local and global feature descriptors to perform a content-based image comparison of the digital images with the example image, to automatically rank the digital images with respect to similarity to the example image. A local feature descriptor may represent a portion of the contents of a digital image. A global feature descriptor may represent substantially all of the contents of that digital image. The global feature descriptor may be content based, not keyword based. Intermediate and final classifiers may be used to perform the automatic ranking. Different intermediate classifiers may generate intermediate relevance metrics with respect to different modalities. The final classifier may use results from the intermediate classifiers to produce a final relevance metric for the digital images. Other embodiments are described and claimed.

Abstract: An image retrieval program (IRP) may be used to query a collection of digital images. The IRP may include a mining module to use local and global feature descriptors to automatically rank the digital images in the collection with respect to similarity to a user-selected positive example. Each local feature descriptor may represent a portion of an image based on a division of that image into multiple portions. Each global feature descriptor may represent an image as a whole. A user interface module of the IRP may receive input that identifies an image as the positive example. The user interface module may also present images from the collection in a user interface in a ranked order with respect to similarity to the positive example, based on results of the mining module. Query concepts may be saved and reused. Other embodiments are described and claimed.

Abstract: An image retrieval program (IRP) may be used to query a collection of digital images. The IRP may include a mining module to use local and global feature descriptors to automatically rank the digital images in the collection with respect to similarity to a user-selected positive example. Each local feature descriptor may represent a portion of an image based on a division of that image into multiple portions. Each global feature descriptor may represent an image as a whole. A user interface module of the IRP may receive input that identifies an image as the positive example. The user interface module may also present images from the collection in a user interface in a ranked order with respect to similarity to the positive example, based on results of the mining module. Query concepts may be saved and reused. Other embodiments are described and claimed.

Abstract: An image retrieval program (IRP) may be used to query a collection of digital images. The IRP may include a mining module to use local and global feature descriptors to automatically rank the digital images in the collection with respect to similarity to a user-selected positive example. Each local feature descriptor may represent a portion of an image based on a division of that image into multiple portions. Each global feature descriptor may represent an image as a whole. A user interface module of the IRP may receive input that identifies an image as the positive example. The user interface module may also present images from the collection in a user interface in a ranked order with respect to similarity to the positive example, based on results of the mining module. Query concepts may be saved and reused. Other embodiments are described and claimed.

Abstract: A speech recognition method includes several embodiments describing application of support vector machine analysis to a mouth region. Lip position can be accurately determined and used in conjunction with synchronous or asynchronous audio data to enhance speech recognition probabilities.

Abstract: A method, apparatus and system including selecting a phase threshold value, receiving a plurality of sequenced buffers, determining a distance between centers of at least two consecutive histogram bins, comparing the distance with the selected threshold value, and determining major execution phases of an executable process based on the comparison.

Abstract: An image retrieval program (IRP) may be used to query a collection of digital images. The IRP may include a mining module to use local and global feature descriptors to automatically rank the digital images in the collection with respect to similarity to a user-selected positive example. Each local feature descriptor may represent a portion of an image based on a division of that image into multiple portions. Each global feature descriptor may represent an image as a whole. A user interface module of the IRP may receive input that identifies an image as the positive example. The user interface module may also present images from the collection in a user interface in a ranked order with respect to similarity to the positive example, based on results of the mining module. Query concepts may be saved and reused. Other embodiments are described and claimed.

Abstract: Human gestures are detected and/or tracked from a pair of digital video images. The pair of images may be used to provide a set of observation vectors that provide a three dimensional position of a subject's upper body. The likelihood of each observation vector representing an upper body component may be determined. Initialization of the model for detecting and tracking gestures may include a set of assumptions regarding the initial position of the subject in a set of foreground observation vectors.

Abstract: A speech recognition method includes use of synchronous or asynchronous audio and a video data to enhance speech recognition probabilities. A two stream factorial hidden Markov model is trained and used to identify speech. At least one stream is derived from audio data and a second stream is derived from mouth pattern data. Gestural or other suitable data streams can optionally be combined to reduce speech recognition error rates in noisy environments.

Abstract: A pattern recognition procedure forms a hierarchical statistical model using a hidden Markov model and a coupled hidden Markov model. The hierarchical statistical model supports a pa 20 layer having multiple supernodes and a child layer having multiple nodes associated with each supernode of the parent layer. After training, the hierarchical statistical model uses observation vectors extracted from a data set to find a substantially optimal state sequence segmentation.

Abstract: An image processing system useful for facial recognition and security identification obtains an array of observation vectors from a facial image to be identified. A Viterbi algorithm is applied to the observation vectors given the parameters of a hierarchical statistical model for each object, and a face is identified by finding a highest matching score between an observation sequence and the hierarchical statistical model.

Abstract: A speech recognition method includes use of synchronous or asynchronous audio and a video data to enhance speech recognition probabilities. A two stream coupled hidden Markov model is trained and used to identify speech. At least one stream is derived from audio data and a second stream is derived from mouth pattern data. Gestural or other suitable data streams can optionally be combined to reduce speech recognition error rates in noisy environments.

Abstract: An arrangement is provided for embedded coupled hidden Markov model. To train an embedded coupled hidden Markov model, training data is first segmented into uniform segments at different layers of the embedded coupled hidden Markov model. At each layer, a uniform segment corresponds to a state of a coupled hidden Markov model at that layer. An optimal segmentation is generated at the lower layer based on the uniform segmentation and is then used to update parameters of models associated with the states of coupled hidden Markov models at lower layer. The updated model parameters at the lower layer are then used to update the model parameters associated with states at the super layer.

Abstract: Human gestures are detected and/or tracked from a pair of digital video images. The pair of images may be used to provide a set of observation vectors that provide a three dimensional position of a subject's upper body. The likelihood of each observation vector representing an upper body component may be determined. Initialization of the model for detecting and tracking gestures may include a set of assumptions regarding the initial position of the subject in a set of foreground observation vectors.

Abstract: A pattern recognition procedure forms a hierarchical statistical model using a hidden Markov model and a coupled hidden Markov model. The hierarchical statistical model supports a pa20 layer having multiple supernodes and a child layer having multiple nodes associated with each supernode of the parent layer. After training, the hierarchical statistical model uses observation vectors extracted from a data set to find a substantially optimal state sequence segmentation.

Abstract: An image processing system useful for facial recognition and security identification obtains an array of observation vectors from a facial image to be identified. A Viterbi algorithm is applied to the observation vectors given the parameters of a hierarchical statistical model for each object, and a face is identified by finding a highest matching score between an observation sequence and the hierarchical statistical model.

Abstract: An arrangement is provided for embedded coupled hidden Markov model. To train an embedded coupled hidden Markov model, training data is first segmented into uniform segments at different layers of the embedded coupled hidden Markov model. At each layer, a uniform segment corresponds to a state of a coupled hidden Markov model at that layer. An optimal segmentation is generated at the lower layer based on the uniform segmentation and is then used to update parameters of models associated with the states of coupled hidden Markov models at lower layer. The updated model parameters at the lower layer are then used to update the model parameters associated with states at the super layer.

Abstract: A visual feature extraction method includes application of multiclass linear discriminant analysis to the mouth region. Lip position can be accurately determined and used in conjunction with synchronous or asynchronous audio data to enhance speech recognition probabilities.

Abstract: A speech recognition method includes use of synchronous or asynchronous audio and a video data to enhance speech recognition probabilities. A two stream coupled hidden Markov model is trained and used to identify speech. At least one stream is derived from audio data and a second stream is derived from mouth pattern data. Gestural or other suitable data streams can optionally be combined to reduce speech recognition error rates in noisy environments.

Abstract: A speech recognition method includes use of synchronous or asynchronous audio and a video data to enhance speech recognition probabilities. A two stream factorial hidden Markov model is trained and used to identify speech. At least one stream is derived from audio data and a second stream is derived from mouth pattern data. Gestural or other suitable data streams can optionally be combined to reduce speech recognition error rates in noisy environments.