A.I. Dev : Blitz Your Life

Hi All, 

Today we will talk about Tim Shaw. A lot of football fans out there might recognize him. In his sixth grade, when Tim won his first trophy, his coach said to his parents, "This kid would go a long way.". But how this way would end, no one would even have had guessed at.

Tim Shaw - The ALS Warrior
Image Courtesy - www.google.com

Since his first trophy, Tim took it seriously, from doing pushups in his room to being the first one at practice, Tim definitely was taking it seriously. He had defined his life with three letters 'N.F.L'. However, soon disaster struck...

After being promoted to the NFL, suddenly his body started responding in ways in which he would never have had made decisions, even in his game. His performance dropped so much that the Titans cut him out. Then after 6 months of time, a doctor finally told him what he was going through and what he should expect from his life hear on. The three letters that the doctor told him were the not the three letters which wandered in his dreams, 'A-L-S', short for Amyotrophic Lateral Schlerosis. And yes, this is indeed as terrible as it sounds, it basically causes the death of neurons controlling the voluntary muscles. So, basically, he wasn't even able to talk properly. He had to think twice every time before moving a single muscle. Once in his talk, he said that a few months after his realization, his phone was not able to understand what he said when he said, "Call Dad". And we have all felt it, how irritating it is when someone just doesn't understand what we are trying to imply.

But help was on its way or at least, kinda.....

Julie Cattiau, a product manager at Google who was working on project Euphonia was trying to work on something similar to Tim's condition. Project Euphonia has two grounds of work, improve speech recognition for people having a variety of medical conditions and the other to give people their voice back. They try to recreate how these people originally used to sound i.e. before they were diagnosed because for a lot of people their voice is their identity. But if you want to recreate a voice, especially of someone who has been diagnosed, the difficult part comes in the beginning. How do you understand them???

Project Euphonia
Image Courtesy - www.google.com

This brings in Speech Recognition. How does this work? let me explain. First off, the sound produced by us is converted into a waveform, you can say it as a picture of our sound. Then this picture is matched up with waveforms or pictures of words already stored in the database. And this is where Machine Learning takes over. Millions of samples are taken into consideration to match the input sounds to output words. Then the algorithm uses Grammar rules, such as syntax, subject-verb agreement, and predicts words to fill up the sentence. This is how one, especially an A.I. can tell the difference between there, their, and they're. But his only works for normal voices. But when it comes to diagnosed people, the going does really get tough. For them, we have to collect voice samples for individual people.

However, the problem is that you do not expect every person to upload around 15,000 samples, so if we have to do, we have to make do with whatever we have. Now, Tim here, with his inspiring inspiration to help out, recorded 2000 samples, but how do you work on them. So initially, the idea is to give the existing recognizer a sentence, and then as and when it fails, we toggle about with its parameters till...Boom! the recognizer is perfectly suitable. Then you just have to make it to read the sentence out loud in a voice that already belonged to that someone. Also known as Voice Synthesis.

Luckily, Tim Shaw being a football star, there were loads of voice samples before he was diagnosed. So indeed luckily, his voice was not totally lost. 

Then as we put in piece and piece, we do have a speech imitator which took in his sentences and made them sound like it was him. A part that was long forgotten by Tim and his family.

...hope you enjoyed this long read, but next time you say 'Ok Google' you just might know how it works!!

#Bringing joy in Everyone's life _ A.I.

@WePeople - signing out for today...

See you with a new post on a new day, goodbye.

P.S.:- 
All expressions in this post are copyrighted to the author (Hriday Kharpude ©).