Detecting self-generated wake expressions

A method comprising:

• producing, by a speaker included in an audio device, output audio in a user environment;
• receiving, from a microphone array of the audio device, a plurality of audio signals representing input audio;
• generating, for each of the plurality of audio signals, a recognition parameter indicating whether a corresponding audio signal includes a predefined expression;
• identifying, based at least in part on the recognition parameters, two or more audio signals, of the plurality of audio signals, that include the predefined expression;
• determining that a total number of the two or more audio signals is more than a threshold number;
• and
• based at least in part on the determining that the total number of the two or more audio signals is more than the threshold number, determining that an occurrence of the predefined expression in the input audio is a result of the predefined expression occurring in the output audio from the speaker.

An audio device comprising:

• one or more processors;
• a plurality of microphones;
• an audio speaker configured to produce output audio;
• memory storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform acts comprising:
• receiving, from the plurality of microphones, a plurality of audio signals representing input audio;
• generating, by the one or more processors for each of the plurality of audio signals, an indication indicating whether a corresponding audio signal contains a predefined expression;
• generating a parameter that indicates at least one of whether or not the output audio is currently being produced, whether or not the output audio contains speech, loudness of the output audio, or an echo characteristic of one or more of the plurality of audio signals;
• identifying, based at least in part on the indications, two or more audio signals, of the plurality of audio signals, that include the predefined expression;
• determining, based at least in part on the two or more audio signals and the parameter, that an occurrence of the predefined expression in the input audio is a result of an utterance of the predefined expression by a user;
• and
• in response to the determining, providing a wake notification corresponding to the predefined expression to the one or more processors to change a state of the audio device.