| 0:00:04 | representing on the speech therapy i bad at that we work and in the fall |
|---|
| 0:00:09 | of two thousand eleven it really is targeted towards children between the three to five |
|---|
| 0:00:14 | one with a cleft lip or pilot which is basically a notch in the upper |
|---|
| 0:00:18 | lip or in the room for them out and |
|---|
| 0:00:22 | one of that side effects of having that is that pronouncing sounds is a great |
|---|
| 0:00:26 | difficulty for them and possible procedures are surgical procedures a prosthetic and speech therapy which |
|---|
| 0:00:33 | is where we come in a rap out is designed after i'd how to train |
|---|
| 0:00:37 | speech therapist would deal with this in which |
|---|
| 0:00:41 | in which i word is prompted in the child is required to repeat that word |
|---|
| 0:00:45 | and the speech their pyjamas where that's right around and this process continues to iteration |
|---|
| 0:00:50 | for different words and phonemes until the correct pronunciations found these were the existing to |
|---|
| 0:00:54 | have to be looked when we were making is a wrap a does something that |
|---|
| 0:00:58 | is more than these |
|---|
| 0:00:59 | because it provides unsupervised feedback and so the child can do it without the strings |
|---|
| 0:01:04 | for our purpose there and that's a good supplement existing work correctly looking like sars |
|---|
| 0:01:09 | they say incorrectly the mlp they will be judged to evaluate harshly there's |
|---|
| 0:01:13 | it is no nothing on the screen that's like is that the strong it'll just |
|---|
| 0:01:16 | keep going we didn't wanna have a negative feedback is really wants to discourage get |
|---|
| 0:01:20 | some try at the end it'll give i get like a little like celebration the |
|---|
| 0:01:24 | results were like avatar that'll dance around for the speech therapist for the |
|---|
| 0:01:27 | for the parent they can also review statistics on how well the channel is doing |
|---|
| 0:01:31 | and how well charleston for specific words even so they can look at worse a |
|---|
| 0:01:35 | specific sounds in it don't know where the target sound or in the word because |
|---|
| 0:01:39 | that's actually also very important for |
|---|
| 0:01:42 | although you a nice be certain things depending on where the sound is maybe more |
|---|
| 0:01:45 | or less people said that only ready us about that longer pronouncing size more accurate |
|---|
| 0:01:51 | estimate |
|---|
| 0:01:52 | and so we bailed five different models for comparison of that was all the pronunciations |
|---|
| 0:01:58 | and we always what we recorded sound we transform it into wave like this with |
|---|
| 0:02:02 | this time |
|---|
| 0:02:04 | so some the signal we can compare it to some baseline a and doubt we |
|---|
| 0:02:08 | have recorded data correct pronunciations about sixty words that are normally most problematic and doubt |
|---|
| 0:02:15 | that not in the at we are able to from but it for that work |
|---|
| 0:02:19 | that judge whether it was set correctly or not |
|---|
| 0:02:23 | so at the end of a day it's you know this may be able to |
|---|
| 0:02:26 | sit down and applying either applying all a computer science as skills that we've had |
|---|
| 0:02:31 | not taking all the classes that were taken in gettysburg and trying to change the |
|---|
| 0:02:35 | world |
|---|