Distributed architecture for training a speech recognition system

摘要

The processing system includes a speech recognizer residing on a computer device. The recognizer receives speech training data and processes the training data into an intermediate representation of the training data. The speech recognizer further communicates the intermediate representation to a second computing device. A speech model server resides on the second computing device. The second computing device is interconnected via a network to the first computing device. The speech model server receives the intermediate representation of the speech training data and generates a speech reference model using the intermediate representation. The speech model server is further operative to communicate the speech reference model to the first computing device. A lexicon coupled to the speech recognizer stores the speech reference model on the first computing device. Independent claims are also included for the following: (a) a distributed speech processing system. (b) a method of building speech reference models for use in a speech recognition system.