Method and apparatus including parallell processes for voice recognition

A system comprising:

• one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
• defining, in an automated speech recognizer in which audio data is processed by a signal conditioning stage followed by a noise suppression stage followed by a language modeling stage, the signal conditioning stage including quantity i processing alternatives, the noise suppression stage including quantity j processing alternatives, and the language modeling stage including quantity k processing alternatives, quantity (i*j*k) alternative paths for processing the audio data through the multiple stages of the automated speech recognizer, i, j, and k being greater than one;
• generating, for each of the quantity (i*j*k) alternative paths, a transcription of particular audio data based on processing the particular audio data through each of the stages of the automated speech recognizer according to the alternative path;
• and
• selecting a particular transcription from among the respective transcriptions that are generated for the quantity (i*j*k) alternative paths signal conditioning stage;
• and
• providing the particular transcription for output.

A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:

• defining, in an automated speech recognizer in which audio data is processed by a signal conditioning stage followed by a noise suppression stage followed by a language modeling stage, the signal conditioning stage including quantity i processing alternatives, the noise suppression stage including quantity j processing alternatives, and the language modeling stage including quantity k processing alternatives, quantity (i*j*k) alternative paths for processing the audio data through the multiple stages of the automated speech recognizer, i, j, and k being greater than one;
• generating, for each of the quantity (i*j*k) alternative paths, a transcription of particular audio data based on processing the particular audio data through each of the stages of the automated speech recognizer according to the alternative path;
• and
• selecting a particular transcription from among the respective transcriptions that are generated for the quantity (i*j*k) alternative paths;
• and
• providing the particular transcription for output.