CPC G10L 15/1807 (2013.01) [G10L 15/063 (2013.01); G10L 21/0264 (2013.01); G10L 25/84 (2013.01); G10L 15/20 (2013.01); G10L 2015/088 (2013.01)] | 14 Claims |
1. A method of training a trigger phrase model, the method comprising:
during a trigger phrase enrollment process:
receiving, at a speech recognition-enabled electronic device associated a user, audio corresponding to the user speaking a trigger phrase; and
based on a count of a number of frames in the audio that have a measure of noise variability of background noise exceeding a noise variability threshold satisfying a threshold value, training, by the speech recognition-enabled electronic device, the trigger phrase model to both:
adapt to a voice of the user of the speech recognition-enabled device using the audio corresponding to the user speaking the trigger phrase; and
detect the trigger phrase in utterances spoken by the user using the audio corresponding to the user speaking the trigger phrase,
wherein the speech recognition-enabled electronic device, while in a sleep mode, is configured to use the trigger phrase model trained during the trigger phrase enrollment process to:
reject the trigger phrase when spoken in utterances by people other than the user of the speech recognition-enabled electronic device; and
wake from the sleep mode when the trigger phrase is spoken in utterances by the user of the speech recognition-enabled electronic device, the sleep mode comprising a power-saving mode of operation in which one or more parts of the speech recognition-enabled electronic device are in a low-power state or powered off.
|