Input device activation noise suppression
A method includes receiving sound input features representative of sound received during an electronic conference, the sound including voice and input device activation sound, receiving an input event feature indicative of the input device activation, and processing the received sound input features and input event feature via a trained model to identify a stored spectral file to be subtracted from the received sound to suppress the input device activation sound.
1. A method comprising:
receiving sound input features representative of sound received during an electronic conference, the sound including voice and input device activation sound;
receiving an input event feature indicative of the input device activation; and
processing the received sound input features and input event feature via a trained model to identify a stored spectral file to be subtracted from the received sound to suppress the input device activation sound wherein the sound input features comprise representations of audio power in multiple frequency bands and the stored spectral file comprises representations of audio power in the same multiple frequency bands.
2. The method of claim 1 wherein the input event feature comprises an indication of a keystroke.
3. The method of claim 2 wherein the input event feature comprises a keyboard identifier.
4. The method of claim 1 wherein the model comprises a deep neural network trained on sound input training features representative of sound received during an electronic conference, the sound including voice and input device activation sound and input activation features.
5. The method of claim 1 wherein the stored spectral file comprises sound represented in the time domain or frequency domain.
6. The method of claim 1 wherein the stored spectral file comprises a wavelet.
7. The method of claim 1 and further comprising:
subtracting the spectral file from the received sound to produce a denoised output; and
transmitting the denoised output in the electronic conference.
8. The method of claim 1 and further comprising transmitting the identified spectral file and the received sound.
9. The method of claim 1 and further comprising:
providing the received sound input features to a voice activity detection model; and
muting transmission of sound in response to no voice activity being detected.
10. The method of claim 1 wherein the input event feature does not identify a specific input from which content can be identified.
11. The method of claim 1 wherein the spectral file is one of multiple spectral files.
12. The method of claim 11 wherein the spectral files are stored on a device driver or device memory.
13. A machine-readable storage device having instructions for execution by a processor of a machine to cause the processor to perform operations to perform a method, the operations comprising:
receiving sound input features representative of sound received during an electronic conference, the sound including voice and input device activation sound;
receiving an input event feature indicative of the input device activation; and
processing the received sound input features and input event feature via a trained model to identify a stored spectral file to be subtracted from the received sound to suppress the input device activation sound wherein the sound input features comprise representations of audio power in multiple frequency bands and the stored spectral file comprises representations of audio power in the same multiple frequency bands.
14. The device of claim 13 wherein the input event feature comprises an indication of a keystroke or a keyboard identifier.
15. A device comprising:
a processor; and
a memory device coupled to the processor and having a program stored thereon for execution by the processor to perform operations comprising:
receiving sound input features representative of sound received during an electronic conference, the sound including voice and input device activation sound;
receiving an input event feature indicative of the input device activation; and
processing the received sound input features and input event feature via a trained model to identify a stored spectral file to be subtracted from the received sound to suppress the input device activation sound wherein the sound input features comprise representations of audio power in multiple frequency bands and the stored spectral file comprises representations of audio power in the same multiple frequency bands.
16. The device of claim 15 wherein the input event feature comprises an indication of a keystroke or a keyboard identifier.
17. The device of claim 15 wherein the model comprises a deep neural network trained on sound input training features representative of sound received during an electronic conference, the sound including voice and input device activation sound and input activation features.
18. The device of claim 15 wherein the operations further comprise
subtracting the spectral file from the received sound to produce a denoised output; and
transmitting the denoised output in the electronic conference.
19. The device of claim 15 wherein the operations further comprise transmitting the identified spectral file and the received sound.