Speech-Selective Audio Mixing for Conference
AssigneePolycom, Inc.
InventorsPeter L. CHU, Yibo LIU, Uri AVNI, Ami NOY
Life of this patent
Estimating the term…
Abstract
A conference apparatus reduces or eliminates noise in audio for endpoints in a conference. Endpoints in the conference are designated as a primary talker and as secondary talkers. Audio for the endpoints is processed with speech detectors to characterize the audio as speech or not and to determine energy levels of the audio. As the audio is written to buffers and then read from the buffers, decisions for the gain settings of faders for read audio of the endpoints being combined in the speech selective mix. In addition, the conference apparatus can mitigate the effects of a possible speech collision that may occur during the conference between endpoints.
Claims33
A method for conducting a conference, comprising:
- buffering audio of each of a plurality of endpoints in the conference with an audio delay;
- leveling the audio of each of the endpoints in the conference with a fader;
- detecting speech in the audio of any one of the endpoints in the conference;
- controlling the audio delay and the fader for each of the endpoints based on the detection of the speech;
- and
- outputting a mix of the audio of the endpoints in the conference based on the control.
An apparatus for conducting a conference, comprising:
- communication equipment connecting a plurality of endpoints in the conference, the communication equipment sending and receiving audio;
- and
- memory having buffers for buffering the audio;
- processing equipment in communication with the communication equipment and the memory, the processing equipment configured to:
- buffer the audio in the buffers of each of the endpoints with an audio delay;
- level the audio of each of the endpoints in the conference with a fader;
- detect speech in the audio of any one of the endpoints;
- control the audio delay and the fader for each of the endpoints based on the detection of the speech;
- and
- output a mix of the audio of each of the endpoints in the conference based on the control.
Dependencies are inferred from the claim text.