OT: Speech Engine

[sym]

Are there any speech experts in this forum?

I know speech domain requires years of experience and is not for faint hearted. This said I am looking into the possibility of building a prototype to support rudimentary english/spanish speech recognition models using open source projects.

Currently I am aware about sphinx, and am wondering if you had any other project suggestions I should consider for speech models and recognition. Thanks for your time.

[desmond]

I have been married over 25 years and the talking never ends, so I consider myself a expert……..

[Allan from Fallbrook]

[quote=desmond]I have been married over 25 years and the talking never ends, so I consider myself a expert……..[/quote]

Desmond: You are a brave, brave man. Perhaps foolhardy, too, but you definitely got you some onions to post that. Sure hope the missus doesn’t catch that one…

[sym]

[quote=desmond]I have been married over 25 years and the talking never ends, so I consider myself a expert……..[/quote]

Very funny 😉

[barnaby33]

Is your model to go from spoken English -> Spanish and vice versa? Speech is highly specialized and from my limited work in it a few years ago, gets very complex very quickly. In my mind the best way to proceed would be to use separate speech to text engines then use something like google translate to do a language text translation.
Josh

[sym]

Josh,

I intend to use separate language models to allow for speech to text, and also to capture some user details from speech/voice patterns. This project is like a crash course on speech for me…. projects like sphinx appear to be language independent speech engines, and one needs to layer the desired language models for the application purpose.

Trying to figure out the necessary modules to build an open source based prototype. Please let me know if you have any recommendations. The closest I worked in this domain was when I intern’ed for a telecommunications company in early 90’s. The application was more text-to-speech synthesize for a personalized message hub, using tcl/tk, dragon dictate and some proprietary algorithms.

Thanks for your help. Cheers.

[CA renter]

I am totally tech ignorant, and not sure if this is in anyway related to what you’re talking about, but the Rosetta Stone language software might have some components that could be somewhat related.

http://www.rosettastone.com/

Sorry if this isn’t helpful, just thinking about language/speech recognition software and decided to post just in case you weren’t familiar with it.

Either way, best of luck to you! Sounds like an interesting project.

[ocrenter]

seems like Nuance has a monopoly on speech recognition and doing a horrible job of it.

would be nice if there will be a viable open source competition.

[livinincali]

I’ve always thought Artificial Neural Networks would be the way to go once processing power caught up. Looks like that might be happening now.

http://research.microsoft.com/en-us/news/features/speechrecognition-082911.aspx

[sym]

[quote=ocrenter]seems like Nuance has a monopoly on speech recognition and doing a horrible job of it.

would be nice if there will be a viable open source competition.[/quote]

That seems to be the case. One or two companies/products I stumbled upon are now part of Nuance. I was even shocked when my favorite editor, Swype, was acquired by them last year.

I am looking for interesting university projects that might be helpful. I also plan to look into Asterisk for possible solutions.

Thx for the suggestion CAR.

[all]

Did you try asking the question on stackoverflow or slashdot?

Edit: This might be good starting point

[Author]

Posts