Back to results
US 12,706,087 B2Granted
Methods and systems for voice control
AssigneeComcast Cable Communications, LLC
InventorsScott Kurtz, Philip Stick, Gary Skrabutenas, Christian Buchter
Life of this patent
Estimating the term…
Abstract
One or more portions of audio input may be detected. Timing data associated with the one or more portions of audio may be determined. Audio processing may be carried out based on the timing data.
Claims20
A method comprising:
- receiving, a first portion of audio data and a second portion of audio data;
- determining, based on the first portion of audio data and the second portion of audio data, a first wake word confidence score;
- determining, based on the second portion of audio data, a second wake word confidence score;
- determining, based on the second wake word confidence score being less than the first wake word confidence score, that the first portion of audio data comprises a beginning of a wake word;
- and
- based on the first portion of audio data comprising the beginning of the wake word, sending the first portion of audio data for processing and excluding the second portion of audio data.
3 more elements
A method comprising:
- receiving a plurality of portions of audio data comprising a wake word;
- determining, based on a first portion of audio data of the plurality of portions of audio data and a second portion of audio data of the plurality of portions of audio data, a first wake word confidence score;
- determining, based on the second portion of audio data and a third portion of audio data of the plurality of portions of audio data, a second wake word confidence score;
- determining, based on the second wake word confidence score being higher than the first wake word confidence score, that a time stamp associated with the second portion of audio data corresponds to a beginning of the wake word;
- and
- based on determining the time stamp associated with the second portion of audio data corresponds to the beginning of the wake word, opening a communication channel with a service.
3 more elements
A method comprising:
- receiving an audio input comprising a plurality of audio frames;
- determining, for each audio frame of the plurality of audio frames, a confidence score that an audio frame of the plurality of audio frames comprises a portion of a wake word;
- determining an increase in the confidence score between a first two audio frames of the plurality of audio frames and a decrease in the confidence score between a second two audio frames of the plurality of audio frames;
- determining, based on the increase in the confidence score, a first boundary of the wake word, and, based on the decrease in the confidence score, a second boundary of the wake word;
- and
- sending only the audio input between the first boundary and the second boundary to a service.
3 more elements
Dependencies are inferred from the claim text.