Author Topic: Speech Recognition  (Read 325 times)

0 Members and 1 Guest are viewing this topic.

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 8264
  • Country: us
Speech Recognition
« on: June 15, 2021, 01:18:55 am »
Many years ago I had an opportunity to tour Bell Labs.  I had friends who worked at Western Electric and had toured parts of their facility as well.  One of the things I remember was seeing a speaker independent recognition system they were working on.   We took turns talking into the microphone and that system was very accurate. Their vision was to allow you to make hands free calls and one of the early applications was to automate the university enrollment over the phone.   Seeing that system work sparked an interest in the subject.  At one point I attempted to make my own system from scratch.   It somewhat works but is very crude.   Later in my career I was on a team working with one of the world leaders on recognition.  In the 90's I had a system from Dragon and an earlier unit that I no longer remember the name.   

Now it's mid 2021 and I have not thought about speech recognition in many years.   I was on a Teams conference the other day and noticed it had a recognition system so I enabled it.  I couldn't believe what I was seeing that Microsoft system do.  Multiple people talking fast and that thing was hardly missing a beat.   There's no way they pulled that out of their ass, so I did a little reading.  A familiar name came up.  Did they buy Dragon or just license the technology.   

What a difference from my home made system from my youth.


How electrically robust is your meter?? https://www.youtube.com/channel/UCsK99WXk9VhcghnAauTBsbg
Software, documentation and test reports for the low cost NanoVNA & V2 Plus 4 may be found here:
https://github.com/joeqsmith
 

Offline evb149

  • Super Contributor
  • ***
  • Posts: 1927
  • Country: us
Re: Speech Recognition
« Reply #1 on: June 15, 2021, 05:59:34 am »
There has been a lot of work going on concerning AI/ML based ASR (automatic speech recognition) via ML (machine learning) using the kinds of neural networks and algorithms that have become dominate in the recent years.  I believe those are different than some algorithms that would have been used in decades past so there's probably some discontinuity in the technology architecture.
Also I think there's a lot that has been going on commercially and in academia that is likely 100% independent of anything dragon did.

You can use the SDKs / services from some of these companies with developer tools and play with the "latest and greatest" yourself.
Nvidia, microsoft, amazon, google, probably all have some offerings in this space.

e.g.

https://developer.nvidia.com/blog/conversational-ai-and-nlp-top-resources-from-gtc-21/

https://developer.nvidia.com/blog/accelerating-conversational-ai-research-with-new-cutting-edge-neural-networks-and-features-from-nemo-1-0/

https://developer.nvidia.com/nvidia-jarvis

https://ngc.nvidia.com/catalog/collections/nvidia:nemo_asr

https://github.com/NVIDIA/NeMo



« Last Edit: June 15, 2021, 06:08:51 am by evb149 »
 

Offline Kerlin

  • Regular Contributor
  • *
  • Posts: 145
  • Country: au
Re: Speech Recognition
« Reply #2 on: June 15, 2021, 06:35:43 am »
I remember the early days of speech recognition.
At the time I read in a technical journal that one bright spark, who was invited to seminar on it, walked in saw on a projector screen that the system was operational and the guests were welcome to try it.
He yelled out "Start, Run, CMD.exe, C, D, C, colon, back slash, enter, format C " and it worked.
« Last Edit: June 15, 2021, 06:49:12 am by Kerlin »
 

Offline Cyberdragon

  • Super Contributor
  • ***
  • Posts: 2599
  • Country: us
Re: Speech Recognition
« Reply #3 on: June 15, 2021, 04:49:35 pm »
I remember the early days of speech recognition.
At the time I read in a technical journal that one bright spark, who was invited to seminar on it, walked in saw on a projector screen that the system was operational and the guests were welcome to try it.
He yelled out "Start, Run, CMD.exe, C, D, C, colon, back slash, enter, format C " and it worked.

That mad lad, he basically ordered the computer to self destruct! :wtf:
*BZZZZZZAAAAAP*
Voltamort strikes again!
Explodingus - someone who frequently causes accidental explosions
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 16214
  • Country: us
Re: Speech Recognition
« Reply #4 on: June 15, 2021, 04:54:56 pm »
IIRC Windows won't let you format the system drive, although maybe there's a way to override that? I never tried *that* hard. It will work on Linux though, sudo rm -r is very dangerous, always, always make sure you are in the directory you intend to be in before using that command.  :palm:
 
The following users thanked this post: SeanB

Offline Kerlin

  • Regular Contributor
  • *
  • Posts: 145
  • Country: au
Re: Speech Recognition
« Reply #5 on: June 15, 2021, 10:28:02 pm »
I remember the early days of speech recognition.

IIRC Windows won't let you format the system drive, although maybe there's a way to override that? I never tried *that* hard.

Zero for reading comprehension.
Thats around 1998.
« Last Edit: June 17, 2021, 10:49:08 pm by Kerlin »
 

Offline Ed.Kloonk

  • Super Contributor
  • ***
  • Posts: 2312
  • Country: au
Re: Speech Recognition
« Reply #6 on: June 16, 2021, 12:04:53 am »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf