Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing predictive pre-recording of audio for voice input. In one aspect, a method includes establishing, as input data, state data that references a state of a mobile device and sensor data that is sensed by one or more sensors of the mobile device, applying a rule or a probabilistic model to the input data, inferring, based on applying the rule or the probabilistic model to the input data, that a user of the mobile device is likely to initiate voice input, and invoking one or more functionalities of the mobile device in response to inferring that the user is likely to initiate voice input.

Abstract: The subject matter of this specification can be implemented in, among other things, a computer-implemented method for correcting words in transcribed text including receiving speech audio data from a microphone. The method further includes sending the speech audio data to a transcription system. The method further includes receiving a word lattice transcribed from the speech audio data by the transcription system. The method further includes presenting one or more transcribed words from the word lattice. The method further includes receiving a user selection of at least one of the presented transcribed words. The method further includes presenting one or more alternate words from the word lattice for the selected transcribed word. The method further includes receiving a user selection of at least one of the alternate words. The method further includes replacing the selected transcribed word in the presented transcribed words with the selected alternate word.

Abstract: Implementations include a method, systems, and computer program products for adaptively configuring codecs to transmit encoded data to better utilize available bandwidth of reachable networks. One implementation is a method of transmitting encoded data from a host to a target. The method includes, selecting a communication network reachable by the host, estimating an available bandwidth from the host to the target via the selected communication network, configuring a codec from a plurality of codecs implemented in the host according to the estimated available bandwidth, encoding data using the configured codec, and transmitting the encoded data to the target.

Abstract: A computer-implemented method of multisensory speech detection is disclosed. The method comprises determining an orientation of a mobile device and determining an operating mode of the mobile device based on the orientation of the mobile device. The method further includes identifying speech detection parameters that specify when speech detection begins or ends based on the determined operating mode and detecting speech from a user of the mobile device based on the speech detection parameters.

Abstract: A method and apparatus are provided for learning a model for the appearance of an object while tracking the position of the object in three dimensions. Under embodiments of the present invention, this is achieved by combining a particle filtering technique for tracking the object's position with an expectation-maximization technique for learning the appearance of the object. Two stereo cameras are used to generate data for the learning and tracking.

Abstract: A system and method facilitating speech detection and/or enhancement utilizing audio/video fusion is provided. The present invention fuses audio and video in a probabilistic generative model that implements cross-model, self-supervised learning, enabling rapid adaptation to audio visual data. The system can learn to detect and enhance speech in noise given only a short (e.g., 30 second) sequence of audio-visual data. In addition, it automatically learns to track the lips as they move around in the video.

Abstract: Systems and methods are described that facilitate assisting a user in filling forms. The invention can parse untagged media, such as email, etc., to identify elements that can be employed to populate a form. Identified elements can be automatically inserted into corresponding fields in a form, and status indicators can be associated therewith and indicated to a user in order to alert the user that review of a particular field in the form is warranted. Additionally, the untagged media and the form can be concurrently presented to the user to facilitate user supervision and/or interaction during assisted form filling.

Abstract: A system and process for controlling common information displays, referred to as shared displays, is presented. The system and process allows multiple modes of input using a set of modules that accept and display data from a variety of sources. Input modules are able to understand data from a single mode of communication and to be able to generate messages as output accordingly. An optional translation module takes discrete message units and converts them into commands or requests that can be processed by a logic module. The logic module includes any application that is running on the shared display. A layout module lays out the information output by the logic module and a display module takes the layout data and converts the information to a form that can be readily displayed on a display device.

Abstract: A browsing system and method for browsing allows multiple users to access and view hypertext documents on a shared display. A browsing system includes a hypertext document converter configured to convert a component in a hypertext document to include alternate component activation tags. A hypertext display controller controls a display module to display the converted component in the hypertext document. The input processor receives and processes an input signal related to the alternate component activation tag from at least one of the plurality of input devices. The browsing system activates the converted component of the hypertext document upon receiving the input signal.

Abstract: A method and apparatus are provided for learning a model for the appearance of an object while tracking the position of the object in three dimensions. Under embodiments of the present invention, this is achieved by combining a particle filtering technique for tracking the object's position with an expectation-maximization technique for learning the appearance of the object. Two stereo cameras are used to generate data for the learning and tracking.

Abstract: A method and apparatus identify a clean speech signal from a noisy speech signal. The noisy speech signal is converted into frequency values in the frequency domain. The parameters of at least one posterior probability of at least one component of a clean signal value are then determined based on the frequency values. This determination is made without applying a frequency-based filter to the frequency values. The parameters of the posterior probability distribution are then used to estimate a set of frequency values for the clean speech signal. A clean speech signal is then constructed from the estimated set of frequency values.