converting speech to text

[naik_mc]

hello
i'm in final year and it's time to find the final year project. i thought to make a project which is useful for the deaf persons to use the phones. i.e to convert someone's speech into text so that one can understand as i'm more intrested in communications.
for this i think that i would need the ad converter. but i don't know how it would b actually used as the practical knowledge for it is very less.
will the project b feasible for final year
or can someone suggest some similar project i think that something that is being used in U.K,U.S.A in the field of electronics but yet to come in the developing countries.
please give your suggestions

 

[rjvh]

it's already exsisting

in my vista package there is a dictation function try to play a bit with it but still didn't get it working right

you have to read a few test sentince and that gives the reference point for the program

in my opinion there is nothing of electronicks involved excepth the mic pick up and the AD converter

its all programing and reconize patterns in the digital info

so i think this question is probably not gona be answerd here in this forum

anyway good luck with your quest

Robert-Jan

 

[awunde]

where are u going?

 

[rjvh]

where are u going?

It's not of your concern where I go

Robert-Jan

 

[jpanhalt]

As implied by Robert-Jan, text-to-speech has been around for a long time (at least since 1990 on home PCs). Commercial units are quite good, and I have used them. Consumer units are almost as good - maybe just as good, but I have much less experience with them. The advances seem to be in software.

Yesterday, I had an experience communicating with a US government agency. The agency used voice to text conversion, and its interpretations of what I said were perfect. This was over a simple telephone line, but the responses were structured, such as "which state were you born in," "spelling mother's maiden name," etc. Nevertheless, I was impressed with the accuracy for something as unpredictable as spelling a name over a telephone link.

John

 

[naik_mc]

that means it can b only used where the questions or the sentences are predefined. but that must happen in the cases of converting text to speech. as the text should b completely in one reference language

 

[rjvh]

that means it can b only used where the questions or the sentences are predefined. but that must happen in the cases of converting text to speech. as the text should b completely in one reference language

Yes and no I think

the reconition is based on tones and order of it and probably checked in a database for similar ones that have a sertain (text) value

if you will have a language that is different than i think that you only have to change the database as diferent combinations of tones (words) have a different text value in an other language

the reconition of the tones on itself is unchanged

Robert-Jan

 

[naik_mc]

Yes and no I think

the reconition is based on tones and order of it and probably checked in a database for similar ones that have a sertain (text) value

if you will have a language that is different than i think that you only have to change the database as diferent combinations of tones (words) have a different text value in an other language

the reconition of the tones on itself is unchanged

Robert-Jan

thats fine for text to speech but in the case of vice versa i think that every tone is different of every individual.that would make it difficult to identify as each would have different tone with different frequency and different pronouncation.
looking at this i think that this project is not feasible.

 

[rjvh]

thats fine for text to speech but in the case of vice versa i think that every tone is different of every individual.that would make it difficult to identify as each would have different tone with different frequency and different pronouncation.
looking at this i think that this project is not feasible.

Till a certain extend yes but no

as this is also exsisting already

your right that everybody has his own vocal caracteristics but they are all in a certain bandwith of tolerance (otherwise nobody would understand eachother)

so it is posible to write a program that just reverse the data to a certain sound that is in the tolerance range so people put a valeu to it

after that it's just the combinations of tones that create the word that caries a meaning in a language

This is just my logical thought about it

Robert-Jan

 

[Mikebits]

People often annunciate words in a unique manner. Inflections, and accent also come into play. From the speech/text converters that I have played with, you have to train the sw to learn your particular speaking patterns.

I think for most voice mail systems, they use questions that invoke a easily recognized response by the sw. In order to perform a complete speech recognition solution is far from perfected as people speak so differently, and I am just referring to one language.

 

[rjvh]

I think for most voice mail systems, they use questions that invoke a easily recognized response by the sw. In order to perform a complete speech recognition solution is far from perfected as people speak so differently, and I am just referring to one language.

yes it is far from perfect but when i go to the bank in cambodia and take a nr the nr is anounced by a computer and it directs you to the designated counter by telling you

this is not a pre recorded voice as you have systems that work like that also

this is all in the Khmer language and at some stage it is typed in with a keybord

I will check some things out on this as one of my friends employees did work on spoken language recognition by computers when he worked in belgium on the university

But I think they are already pretty far with a working product (but perfect is a high goal to achieve)

Robert-Jan

 

[ericgibbs]

yes it is far from perfect but when i go to the bank in cambodia and take a nr the nr is anounced by a computer and it directs you to the designated counter by telling you

this is not a pre recorded voice as you have systems that work like that also

this is all in the Khmer language and at some stage it is typed in with a keybord

I will check some things out on this as one of my friends employees did work on spoken language recognition by computers when he worked in belgium on the university

But I think they are already pretty far with a working product (but perfect is a high goal to achieve)

Robert-Jan

hi Robert-Jan,
microsoft have masses of voice recognition freeware, both for speaking and control.

I used to control my PC with speech, but I do prefer to use the keyboard.
I made a parallel port interface to the PC, I could play music, change TV channels and call up programs by speech, this was about 3 years ago.

 

[naik_mc]

hi Robert-Jan,
microsoft have masses of voice recognition freeware, both for speaking and control.

I used to control my PC with speech, but I do prefer to use the keyboard.
I made a parallel port interface to the PC, I could play music, change TV channels and call up programs by speech, this was about 3 years ago.

but that's just for a single person. and one can receive a phone call of no. of persons

 

[blueroomelectronics]

Since it's your final year project aren't you supposed to pick something that demonstrates your skills & knowledge?

 

[naik_mc]

Since it's your final year project aren't you supposed to pick something that demonstrates your skills & knowledge?

so what's ur view?

 

[TuTu]

i did that project.
but my advise is ...
DONT do this project...