Speech model personalization via ambient context harvesting
AssigneeIntel Corporation
InventorsGabriel Amores, Guillermo Perez, Moshe Wasserblat, Michael Deisher, Loic Dufresne de Virel
Life of this patent
Estimating the term…
Abstract
An apparatus for speech model with personalization via ambient context harvesting, is described herein. The apparatus includes a microphone, context harvesting module, confidence module, and training module. The context harvesting module is to determine a context associated with the captured audio signals. A confidence module is to determine a confidence of the context as applied to the audio signals. A training module is to train a neural network in response to the confidence being above a predetermined threshold.
Claims25
An apparatus for speech modeling via ambient context harvesting, the apparatus comprising:
- a microphone to capture audio signals;
- processor circuitry;
- context harvesting instructions to cause the processor circuitry to:
- detect speech based on the audio signals;
- identify one or more of a location or a time of day associated with the capture of the audio signals;
- and
- determine a context associated with the audio signals based on the one or more of the location or the time of day;
- confidence analysis instructions to cause the processor circuitry to:
- determine a confidence score indicative of a likelihood that the speech is associated with the context;
- and
- identify the confidence score as satisfying a confidence threshold;
- and
- training instructions to train a neural network based on the audio signals corresponding to the speech after the confidence score satisfies the confidence threshold.
A system comprising:
- a microphone to capture audio signals;
- memory;
- machine-readable instructions:
- and
- processor circuitry to execute the machine-readable instructions to:
- determine a context associated with the audio signals based on one or more of a location or a time of day associated with the capture of the audio signals, the context including an expected speech interaction;
- detect speech based on the audio signals;
- determine a confidence score corresponding to a likelihood that the speech is associated with the expected speech interaction of the context;
- and
- perform a comparison of the confidence score to a confidence threshold;
- and
- classify the audio signals corresponding to the speech as training data to be used to train a neural network based on satisfaction of the confidence threshold by the confidence score.
A method comprising:
- detecting speech based on audio signals captured from an environment;
- predicting a context associated with the captured audio signals based on one or more properties of the environment, the one or more properties different than noise in the environment;
- determining a confidence score indicating a likelihood that the speech is associated with the context;
- performing a comparison of the confidence score to a confidence threshold;
- and
- training a neural network based on the audio signals corresponding to the speech in response to the confidence score satisfying the confidence threshold.
Dependencies are inferred from the claim text.