Transforming voice signals to compensate for effects from a facial covering
AssigneeCisco Technology, Inc.
InventorsEric Y. Chen, Shamim S. Pirzada, Cullen Frishman Jennings
Life of this patent
Estimating the term…
Abstract
In one example embodiment, audio characteristics of audio signals are adjusted by a first machine learning model to reduce effects of a facial covering and produce adjusted audio signals. The audio signals correspond to resulting voice signals produced from the facial covering affecting original voice signals. Speech characteristics are predicted for the adjusted audio signals by a second machine learning model. Transformed audio signals corresponding to the original voice signals are produced based on the adjusted audio signals and predicted speech characteristics.
Claims20
A method comprising:
- adjusting audio characteristics of audio signals by a first machine learning model to reduce effects of a facial covering and produce adjusted audio signals, wherein the audio signals correspond to resulting voice signals produced from the facial covering affecting original voice signals and the first machine learning model determines a first confidence value for the adjusted audio signals;
- predicting phoneme and prosody for the adjusted audio signals by a second machine learning model to produce predicted phoneme and prosody, wherein the second machine learning model determines a second confidence value for the predicted phoneme and prosody;
- determining a natural language of the adjusted audio signals and a third confidence value by a third machine learning model;
- and
- producing transformed audio signals corresponding to the original voice signals in response to the third confidence value satisfying a threshold by combining the adjusted audio signals with the predicted phoneme and prosody based on the first confidence value for the adjusted audio signals and the second confidence value for the predicted phoneme and prosody.
An apparatus comprising:
- a computing system comprising one or more processors, wherein the one or more processors are configured to:
- adjust audio characteristics of audio signals by a first machine learning model to reduce effects of a facial covering and produce adjusted audio signals, wherein the audio signals correspond to resulting voice signals produced from the facial covering affecting original voice signals and the first machine learning model determines a first confidence value for the adjusted audio signals;
- predict phoneme and prosody for the adjusted audio signals by a second machine learning model to produce predicted phoneme and prosody, wherein the second machine learning model determines a second confidence value for the predicted phoneme and prosody;
- determine a natural language of the adjusted audio signals and a third confidence value by a third machine learning model;
- and
- produce transformed audio signals corresponding to the original voice signals in response to the third confidence value satisfying a threshold by combining the adjusted audio signals with the predicted phoneme and prosody based on the first confidence value for the adjusted audio signals and the second confidence value for the predicted phoneme and prosody.
One or more non-transitory computer readable storage media encoded with processing instructions that, when executed by one or more processors, cause the one or more processors to:
- adjust audio characteristics of audio signals by a first machine learning model to reduce effects of a facial covering and produce adjusted audio signals, wherein the audio signals correspond to resulting voice signals produced from the facial covering affecting original voice signals and the first machine learning model determines a first confidence value for the adjusted audio signals;
- predict phoneme and prosody for the adjusted audio signals by a second machine learning model to produce predicted phoneme and prosody, wherein the second machine learning model determines a second confidence value for the predicted phoneme and prosody;
- determine a natural language of the adjusted audio signals and a third confidence value by a third machine learning model;
- and
- produce transformed audio signals corresponding to the original voice signals in response to the third confidence value satisfying a threshold by combining the adjusted audio signals with the predicted phoneme and prosody based on the first confidence value for the adjusted audio signals and the second confidence value for the predicted phoneme and prosody.
Dependencies are inferred from the claim text.