Voice Recognition White Papers
CSR: Speaker Recognition From Compressed VoIP Packet Stream
Overview CELP compressed algorithms (e.g., G729, G723, GSM, etc.) are pervasively used in the emerging VoIP phone calls. It is desired (e.g., for security reason) in many VoIP applications to identify the speaker. This paper investigates an innovative scalable approach to achieve the speaker ID directly from the compressed voice packets. A set of feature vector is first identified and grouped from the compressed voice stream, and then fast clustering mechanism is employed to classify the voice packets for real time speaker identification.
| Publisher | IBM | File Format | PDF, requires Acrobat Rdr 5 |
|---|---|---|---|
| Date Published | January 2005 | Downloads | 1019 |
| Format | White Papers | ||
| Topics | |||



