摘要 |
A method of accelerating a speech recognition system comprising providing a task scheduler for a phoneme based audio processor. The task scheduler for communicating with the elements of the phoneme based audio processor, including but not limited to buffer memory, labeler, digital signal processor, and Viterbi processor, to manage the processing of these elements in relation to performance measures of the speech recognition system. Advantageously the task scheduler allowing for the optimization of power consumption, adjustment of processing to avoid overflow of buffer memory, and allowing the dynamic provisioning of multiple hardware, firmware and software elements of the audio processor.
|