Aircraft speech recognition systems and methods

A speech recognition method in an aircraft, comprising:

• executing, via at least one processor, an air traffic control transcription application using first acoustic and language models, wherein the air traffic control transcription application is configured to recognize a first set of words, receive audio data and to output a transcription of the audio data, the air traffic control transcription application having a first response time;
• executing, via the at least one processor, a command and control speech recognition application using second acoustic and language models, wherein the command and control speech recognition application is different from the air traffic controller transcription application, has a second response time that is faster than the first response time, and is configured to recognize a second set of words and wherein the second set of words includes fewer words than the first set of words;
• receiving, via the at least one processor, flight context data, the flight context data including at least one of transcription text data output from the air traffic control transcription application, flight plan data, current aircraft location data, user interface data, and displays system data;
• performing natural language processing, via the at least one processor, on the transcription to identify additional words including at least one named entity corresponding to a location, wherein the additional words are not included in the second set of words and are useful based on the flight context data;
• extracting, via the at least one processor, acoustic and language model parameters corresponding to the additional words including the at least one named entity from the first acoustic and language models;
• and
• adding, via the at least one processor, the extracted acoustic and language model parameters corresponding to the additional words including the at least one named entity from the first acoustic and language models to the second acoustic and language models to generate an updated second set of words including the additional words;
• determining, via the at least one processor, discardable words in the second set of words based on the flight context data;
• removing, via the at least one processor, acoustic and language model parameters corresponding to the discardable words from the second acoustic and language model thereby removing the discardable words from the second set of words;
• and
• generating, via the at least one processor, an aircraft control command encapsulating at least one of the additional words using the command and control speech recognition application, wherein the at least one of the additional words is recognized by the second acoustic and language models with the extracted acoustic and language model parameters that have been added to the second acoustic and language models.