The output volume of a device is usually controlled as a fixed value between zero to a maximum. Unless manually adjusted by a user, the volume of the device output does not vary. In the case of speech output, such constant-volume responses can sound unnatural because human speech typically involves nuanced variations in volume that serve as subtle communicative signals. This disclosure describes techniques to dynamically and automatically adjust the volume of devices such as smartphones, smart speakers, smart appliances, etc. The adjustment is achieved by determining the appropriate level for output volume such that a fixed signal-to-noise (SNR) ratio is maintained. Further, if the user permits, parameters related to the user’s context can be used to determine adjustment of the SNR to fit the user’s mood. The techniques can be implemented via a trained machine learning model.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.