Parallel recognition

摘要

Recognition techniques may include the following. On a first processing entity, a first recognition process is performed on a first element, where the first recognition process includes: in a first state machine having M (M>1) states, determining a first best path cost in at least a subset of the M states for at least part of the first element. On a second processing entity, a second recognition process is performed on a second element, where the second recognition process includes: in a second state machine having N (N>1) states, determining a second best path cost in at least a subset of the N states for at least part of the second element. At least one of the following is done: (i) passing the first best path cost to the second state machine, or (ii) passing the second best path cost to the first state machine. The foregoing techniques may include one or more of the following features, either alone or in combination.

主权项

1. A method comprising:
obtaining speech data for a particular frame representing a portion of an utterance; providing the speech data to multiple processors that are configured to process the speech data in parallel, each of the multiple processors being assigned to evaluate a different set of path segments in a network of states; before the multiple processors receive speech data for future frames that occur after the particular frame and represent subsequent portions of the utterance, receiving, from the multiple processors, cost values that represent an extent to which path segments match one or more of the future frames not yet received by the multiple processors; communicating, among the multiple processors, the cost values that represent the extent to which path segments match the one or more of the future frames not yet received by the multiple processors; and using the communicated cost values to determine a transcription of at least a portion of the utterance.