摘要 |
A method and system for distributed speech recognition in a communications network which includes at least one speech processing server coupled to a plurality of communication terminals, wherein a front-end device residing at each terminal is used to extract an amount of speech features from voice signals and a back-end device residing at the server to recognize words from the extracted features. The front-end device is configurable so that the amount of speech features can be adjusted based on the prevailing conditions that affect the speech processing. At the server side, a traffic monitoring device is used to determine how many users are accessing the back-end recognizer so that the amount of extracted features can be increased when the channel traffic is light and reduced when the traffic is heavy. |