| 0:00:15 | so a high everyone |
|---|
| 0:00:18 | are all be presenting local binary patterns as features for speaker recognition |
|---|
| 0:00:24 | so basically we are trying to replace the classical delta and that's of the times |
|---|
| 0:00:32 | as you know they are useful to describe the temporal evolution of frames but here |
|---|
| 0:00:38 | we're using a concept used in face recognition |
|---|
| 0:00:42 | it's called a local many buttons and it's useful to describe the whole neighbourhood |
|---|
| 0:00:49 | related to one pixel |
|---|
| 0:00:51 | so it has been used in this |
|---|
| 0:00:54 | not exactly speaker recognition but in i by spoofing and it has been a proved |
|---|
| 0:01:00 | to be quite efficient |
|---|
| 0:01:02 | i guess the artificial voice the extra |
|---|
| 0:01:06 | so we're try |
|---|
| 0:01:10 | we're using it for speaker recognition and along with mfccs and where proving that |
|---|
| 0:01:18 | it's |
|---|
| 0:01:20 | more efficient and mfcc plus data but that senators thank you |
|---|