Cisco Systems 5.3.x manual About Pulse Speaker Identification

Chapter 4 Play, Comment, Tag, and Share Videos

Procedures

About Pulse Speaker Identification

The Pulse engine can identify different speakers within videos. It does this by identifying unique voiceprints.

Speaker identification is a learned process. At first, voiceprints are assigned a generic speaker tag, such as Speaker 1. Authors can then use the Cisco Show and Share editing tools to associate voiceprints with specific speakers. The speaker names are taken from the Cisco Show and Share registered users list.

Authors cannot assign names outside of the user pool.

Once a voiceprint has been associated with a specific user, subsequent videos uploaded with that voiceprint automatically show the speaker name.

The colored areas in the video timeline, below, mark each instance of a speaker in the video. This video has a single speaker—the gray areas between the yellow bars indicate where no one is speaking.

The accuracy of the Pulse speaker identification depends upon the quality of the recorded audio and whether or not there is a lot of background noise in the audio track. Sometimes a single speaker may show up as named and as a generic speaker in the same video because of background noise or changes in recording quality. The more times a speaker is identified by a video author, the more accurate the system becomes.

See Label Unidentified Speakers, page 5-12, for information about how to assign speakers to unidentified voiceprints.

User Guide for Cisco Show and Share 5.3.x

4-19