Dragon Systems, Inc.

Dean G Sturtevant

Joel M Gould

James K Baker

Frank J McGrath

Jed M Roberts

Joel W Parke

Steven D Squires

Peter S Heitman

Elizabeth E Steele

Laurence Steven Gillick

Robert S Roth

Janet M Baker

A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string; score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary state associated with that program unit into which vocabulary words which will be made active when that program group has the focus can be put; detect the available computational resources and alter the instructions it executes in response; test if its ability to respond to voice input has been shut off without user confirmation, and, if so, turn that ability back on and prompt the user to confirm if that ability is to be turned off, store both a first and a second set of models for individual vocabulary words and enable a user to selectively cause the recognizer to disregard the second set of models for a selected word; and/or score a signal representing a given word against models for that word from different word model sets to select which model should be used for future recognition.

Apparatuses and methods for training and operating speech recognition

A word recognition system detects the computational resources available to it, such as the speed or number of processors, or whether there is DSP hardware, and alters the instructions it executes in response. The system can be a word recognition program designed to run on different computers having different computational resources. The program receives user generated word signals representing words to be recognized; performs pattern matching on them to select which vocabulary words most probably correspond to such word signals; detects if certain computational resources are available; and varies the instructions it executes in response. In many embodiments the system is a speech recognition program. The word recognition program can vary the computational intensity of its signal processing as a function of available computational resources. Preferably it can match the same word models against representations of word signals produced both by its more and less intensive signal processing. For example, the more intense signal processing might take more frequent FFTs, but average them over the same time period at which the less intense signal processing takes FFTs, so both can represent word signals with the same data structure. If the program detects both a CPU and a DSP processor, it can cause the DSP to determine when the program should interrupt the CPU. The program can also vary the rate at which it filters relatively low scoring words out of consideration during the recognition process as a function of the level of available computational resources.

Word recognition system which alters code executed as a function of

A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string; score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary state associated with that program unit into which vocabulary words which will be made active when that program group has the focus can be put; detect the available computational resources and alter the instructions it executes in response; test if its ability to respond to voice input has been shut off without user confirmation, and, if so, turn that ability back on and prompt the user to confirm if that ability is to be turned off; store both a first and a second set of models for individual vocabulary words and enable a user to selectively cause the recognizer to disregard the second set of models for a selected word; and/or score a signal representing a given word against models for that word from different word model sets to select which model should be used for future recognition.

A method is provided for deriving acoustic word representations for use in speech recognition. Initial word models are created, each formed of a sequence of acoustic sub-models. The acoustic sub-models from a plurality of word models are clustered, so as to group acoustically similar sub-models from different words, using, for example, the Kullback-Leibler information as a metric of similarity. Then each word is represented by cluster spelling representing the clusters into which its acoustic sub-models were placed by the clustering. Speech recognition is performed by comparing sequences of frames from speech to be recognized against sequences of acoustic models associated with the clusters of the cluster spelling of individual word models. The invention also provides a method for deriving a word representation which involves receiving a first set of frame sequences for a word, using dynamic programming to derive a corresponding initial sequence of probabilistic acoustic sub-models for the word independently of any previously derived acoustic model particular to the word, using dynamic programming to time align each of a second set of frame sequences for the word into a succession of new sub-sequences corresponding to the initial sequence of models, and using these new sub-sequences to calculate new probabilistic sub-models.

Method for representing word models for use in speech recognition

Growing community of inventors

Watertown, MA, United States of America

Dean G Sturtevant