speech recognition for all words by database

Question

0.00/5 (No votes)

See more:

, +

hello everyone, i have done speech recognition app offline by using the speech library which call load grammer method which contains the words in code.

but it is not good if the words written in the code , because the english language is huge so it is good to make databasefor all words

but it is hard, can i store all words in database ?? and how can i make correlation between my spoken word and the saved word in database ??

can you help me, please ?

C#

Choices SLES = new Choices(); // class Choice of loadGrammer Method
            SLES.Add(new String[] {"hello","how","are" ,"you","book","conversation","the","peace","he","translates", "text", "from","Arabic","to","English","language","exit", "everyone", "in","translator","it" });
            Grammar gr = new Grammar(new GrammarBuilder(SLES));
            try
            {
                recognizer.RequestRecognizerUpdate(); //Request of Recognition
                recognizer.LoadGrammar(gr); //load Grammer
                recognizer.SpeechRecognized += recognizer_SpeechRecognized;
                recognizer.SetInputToDefaultAudioDevice(); 
                recognizer.RecognizeAsync(RecognizeMode.Multiple); 

                    }

Posted 20-Nov-14 4:49am

Me is Needer

Updated 4-Aug-21 16:47pm

PIEBALDconsult

v2

Add a Solution

Comments

BillWoodruff 20-Nov-14 10:58am

"can i store all words in database ?" Why not ?

"how can i make correlation between my spoken word and the saved word in database ?" the same way you query a database for a match(es) with any other type of data.

But, do you really need a database here, or do you need some kind of static class full of static information that you compile to a .dll and use in your project ?

Sergey Alexandrovich Kryukov 20-Nov-14 11:26am

There is a dictation grammar already, which can be used via system.speech.dll, but the quality...
The problem is not the dictionary, the problem is: if you have words/phrases even remotely similar in pronunciation, recognition fails to pick the right one. I think doing it well is nearly impossible without detailed semantic natural language recognition, otherwise it will recognize "Nutcracker" as "night cracker" (real-life example!), and other things like that... :-)

Anyway, please see my answer.

—SA

Me is Needer 20-Nov-14 12:09pm

thank you so much , but fisrt my project is simple for isolated word(word by word)

i have used system.speech.dll, but the words to recognize is written in the code, this is my proplem, i dont want the words be in code because i want to use all english words and they are huge

guide me, please

Sergey Alexandrovich Kryukov 20-Nov-14 12:14pm

This is another problem. You can always read those words, for example, from some file...
—SA

Me is Needer 20-Nov-14 12:41pm

Mr Sergey, i have done the above code , those words like"hello,..." are written inside the code this is for simple collection words, but if i want to contain all words in English , how can i do ??

Sergey Alexandrovich Kryukov 20-Nov-14 12:47pm

Please see my answer. Please see my comment to Bill's comment above though.
—SA

[no name] 20-Nov-14 13:58pm

That is correct, reading a file of English words is one thing and easy enough, but the problem with this is that you are then dealing with an entirely different issue thereafter; as BillWoodruff has said, that you would practically need to train your application, which is an entirely different task altogether.

Sergey Alexandrovich Kryukov 20-Nov-14 16:41pm

I've heard that some original speech recognition products (not based on Microsoft Speech API and not licensed to Microsoft for free distribution :-( ) provide reasonable dictation quality, but I never tried out anything of this kind. In all cases, all mass-production systems (such as phone-based call centers) are way too far from such quality (and, actually, are hardly acceptable for what they are already used).

A need for training would greatly compromise the mass customer usage and is not actually used in Microsoft speech recognition...

—SA

[no name] 20-Nov-14 17:10pm

I never looked before for a library for this, but you are right, there are other options available, but I honestly don't see the point reinventing the wheel. :)

Maybe this might be of interest to the OP; it made an interesting read for me: https://software.intel.com/en-us/articles/voice-recognition-and-synthesis-using-the-intel-perceptual-computing-sdk

There are a dozen others just like the one Microsoft offer if you search Google for C# Speech Recognition Library. You will find some of the ones Sergey is probably referring to.

Sergey Alexandrovich Kryukov 21-Nov-14 1:06am

You could post this link as an extra answer...
—SA

[no name] 21-Nov-14 3:57am

Thanks Sergey, done ;)

BillWoodruff 20-Nov-14 12:45pm

In fact, the problem is NP-Hard :)

[no name] 20-Nov-14 11:31am

Creating a DLL project and include it as a dependency of the main application project would be how I would do this too. Storing all the words in a DLL class and simply include the DLLs namespace in your main project. No need for a database here.

The easiest way to add all the words is to download a dictionary and iterate each word and add it to a new text file, something like this: https://raw.githubusercontent.com/eneko/data-repository/master/data/words.txt

Then add that text file as a resource to your DLL project. Job Done

Me is Needer 20-Nov-14 12:38pm

thank you for your answer , but what do you mean by static class ?

BillWoodruff 20-Nov-14 12:47pm

Please forget about the idea of "all English words." Choose a sub-set of ... well: how much time do you, or your users, have to train the app ?

Have you thought about interfacing with a commercial speech recognition program like Dragon ? For reasons I don't care to understand Chant is now providing C# api's to Dragon:

http://www.chant.net/Support/KnowledgeBase/HowTos/H071128.aspx

PIEBALDconsult 20-Nov-14 11:07am

It's an impossible task; leave it to the experts.

3 solutions

Solution 3

Maybe this might be of interest to the OP; it made an interesting read for me: https://software.intel.com/en-us/articles/voice-recognition-and-synthesis-using-the-intel-perceptual-computing-sdk

There are a dozen others just like the one Microsoft offer if you search Google for C# Speech Recognition Library. You will find some of the ones Sergey is probably referring to.

Posted 20-Nov-14 21:56pm

Sheepings

Updated 20-Nov-14 21:58pm

v2

Comments

Sergey Alexandrovich Kryukov 21-Nov-14 11:25am

5ed.
—SA

[no name] 21-Nov-14 15:19pm

Thank you; I also gave you 5 for your solution ;)

Sergey Alexandrovich Kryukov 21-Nov-14 16:31pm

Thank you.
—SA

Solution 2

There are some articles on code project, you may take help from it

Speech recognition, speech to text, text to speech, and speech synthesis in C#[^]

Speech Recognition for the Web[^]

Posted 20-Nov-14 8:15am

Shweta N Mishra

Updated 20-Nov-14 8:17am

v2

Add a Solution

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Accepted Answer · 2014-11-20T05:22:00

There is a predefined grammar for dictation covering some big dictionary, System.Speech.Recognition.DictationGrammar:
http://msdn.microsoft.com/en-us/library/system.speech.recognition.dictationgrammar%28v=vs.110%29.aspx[^].

However, don't hold your breath. I tried it with the engine supplied by Microsoft for Windows XP and 7. The quality of recognition for small grammars with distinctly different words/phrases was fairly good, but for dictation it was quite bad. I'm not sure if you can get something much better. See also my comment to the question.

—SA

speech recognition for all words by database

3 solutions

Solution 1

Solution 3

Solution 2

Add your solution here

Preview 0

speech recognition for all words by database

3 solutions

Solution 1

Solution 3

Solution 2

Add your solution here

Preview 0

Existing Members

...or Join us