Distributed speech recognition using one way communication

摘要

A speech recognition client sends a speech stream and control stream in parallel to a server-side speech recognizer over a network. The network may be an unreliable, low-latency network. The server-side speech recognizer recognizes the speech stream continuously. The speech recognition client receives recognition results from the server-side recognizer in response to requests from the client. The client may remotely reconfigure the state of the server-side recognizer during recognition.

主权项

1. A method performed by at least one computer processor executing computer program instructions stored on at least one non-transitory computer-readable medium, the method comprising:
(A) receiving a speech stream and a control stream from a client, the speech stream including a minimum configuration state identification number required to begin recognition of a first portion of the speech stream from a client; (B) determining whether a configuration state identification number associated with a state of an automatic speech recognition engine is at least as great as the received minimum configuration state identification number; (C) if the configuration state identification number associated with the state of the automatic speech recognition engine is determined to be at least as great as the received minimum configuration state identification number, then using the automatic speech recognition engine to recognize the first portion of the speech stream and thereby to produce a first speech recognition result; and (D) if the configuration state identification number associated with the state of the automatic speech recognition engine is not determined to be at least as great as the received minimum configuration state identification number, then incrementing the configuration state identification number associated with the state of the automatic speech recognition engine until the configuration state identification number associated with the state of the automatic speech recognition engine is determined to be at least as great as the received minimum configuration state identification number before using the automatic speech recognition engine to recognize the first portion of the speech stream and thereby to produce the first speech recognition result.