0:00:11 | hi my name is are free a i'm a computer science phd student graduating then |
---|
0:00:15 | this year |
---|
0:00:15 | and i work at the international |
---|
0:00:19 | i'll give you guys abroad interview of icsi is recall |
---|
0:00:22 | it's a off campus down by the bart station |
---|
0:00:25 | we have multiple groups down there are have an open house |
---|
0:00:29 | where will have a shuttle leaving here also just look for me an orange jacket |
---|
0:00:32 | or just walk down for support set nineteen forty seven centre street |
---|
0:00:37 | so my group in particular dixie is the speech group where you would lead by |
---|
0:00:41 | professor nelson morgan he's the each professor |
---|
0:00:44 | able recently transitioned a task even like men as a leader he's |
---|
0:00:48 | gotta industry background he's been and you want to voice signal |
---|
0:00:52 | dragons this go companies or a well-known in speech recognition field |
---|
0:00:56 | about a half dozen phd students clean myself |
---|
0:01:01 | so our main research areas we have speech recognition |
---|
0:01:04 | this is speech-to-text taking an audio signal and converting it to words it's a field |
---|
0:01:09 | with a long history that's a system from the really nineteen sixties at i b |
---|
0:01:13 | m the recognized just the digits zero to ten |
---|
0:01:16 | later in the nineties and now it's used for dictation in medical nickel confessions |
---|
0:01:21 | and most recently you know it's been adopted in smart phones and digital assistants likes |
---|
0:01:25 | your |
---|
0:01:27 | we also do work in speaker recognition this is sort of voice biometrics |
---|
0:01:31 | and we have a multimedia processing group which works in collaboration with a computer vision |
---|
0:01:35 | researchers |
---|
0:01:36 | both on campus and i dixie |
---|
0:01:40 | so speech recognition this is the main thrust of our groups work |
---|
0:01:44 | one of the project we have is looking at hidden markov models which underlie every |
---|
0:01:48 | single speech recognition system |
---|
0:01:51 | but it's based on a few flawed assumptions so years a graphical model representation of |
---|
0:01:55 | a hidden markov model |
---|
0:01:57 | we have observations and the darker shaded scores above and |
---|
0:02:00 | hidden variables states on the bottom |
---|
0:02:03 | and if you are familiar with the graphical model for lance were saying that all |
---|
0:02:07 | the observations are conditionally independent |
---|
0:02:10 | given the current state |
---|
0:02:11 | i know that a simplifying assumption that makes algorithms tractable |
---|
0:02:15 | but has consequences because reality is that these observations do you have a lot of |
---|
0:02:19 | contextual |
---|
0:02:20 | correlation |
---|
0:02:22 | our work in acoustic features we look at basically dealing |
---|
0:02:25 | with these shaded circles on top on their very noisy |
---|
0:02:28 | you know speech recognition what's great clean conditions when real conditions the data is different |
---|
0:02:32 | there's been a resurgence of interest in neural networks this is sort of a throwback |
---|
0:02:36 | to the nineteen eighties that's been |
---|
0:02:39 | kind of in some excitement because the restricted boltzmann machines and more computational power allows |
---|
0:02:46 | a bill much larger |
---|
0:02:48 | before networks and the research that i do specifically is related to |
---|
0:02:52 | accordingly system that are very complex and somewhat targeted two major european languages |
---|
0:02:57 | to languages i have fewer resources |
---|
0:03:00 | so on this map of language families of the world |
---|
0:03:03 | the parts and blue and red are the major european languages and the rest the |
---|
0:03:07 | world most the stuff on the right |
---|
0:03:10 | i have languages very different characteristics |
---|
0:03:14 | in speaker recognition our work is really its robustness |
---|
0:03:18 | because this work is related to |
---|
0:03:20 | to security this work is funded by the army for the air force research labs |
---|
0:03:25 | and darpa |
---|
0:03:25 | and we're dealing with very difficult signals collected like |
---|
0:03:28 | and |
---|
0:03:30 | jet fighter cockpit and spire planes |
---|
0:03:33 | but there's also another application which is speaker diarization |
---|
0:03:36 | and we've done this in collaboration so far left building a real time online system |
---|
0:03:40 | which given up |
---|
0:03:42 | recording of a meeting say with multiple speakers |
---|
0:03:45 | can |
---|
0:03:46 | segment and label the regions which correspond to different speakers this helps you if you |
---|
0:03:51 | want to build a transcript afterwards and have a labeled with different speakers |
---|
0:03:57 | like the other work we do is related to this in |
---|
0:03:59 | in terms of speech activity detection language identification and these dialogue systems we have speakers |
---|
0:04:06 | who |
---|
0:04:07 | speaker over each other they speak with each other and so we have to deal |
---|
0:04:11 | this differently |
---|
0:04:12 | a lot of it has to do with source speech understanding also rather than just |
---|
0:04:16 | getting the words we try to understand what dynamics are what the intent is behind |
---|
0:04:20 | the language and some of the more exciting work we're doing also involves looking at |
---|
0:04:24 | actual |
---|
0:04:25 | i data collected by scientists who |
---|
0:04:27 | have access to |
---|
0:04:29 | to brain measurements from small mammals or from patients |
---|
0:04:33 | for example they do surgery on epilepsy patients and they stick electrodes quality in the |
---|
0:04:38 | surgery |
---|
0:04:39 | in the range |
---|
0:04:40 | thus circles at icsi are we wanna have more robust signal processing for realistic conditions |
---|
0:04:46 | we want to understand the scientific principles behind these engineering systems |
---|
0:04:52 | and in general we support open collaborative research we work with |
---|
0:04:55 | a lot of international universities and false domestically and i encourage you guys to check |
---|
0:05:00 | up website and company open house today |
---|
0:05:07 | policy of harmony |
---|