Win a copy of Functional Reactive Programming this week in the Other Languages forum!
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic

Can any one Direct me on creating a Speaker Recognition System in Java?

 
Varuna Seneviratna
Ranch Hand
Posts: 170
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
What API is to use to detect the speaker's voice?
What is the fundamental approach to follow in order to identify a persons voice irrespective of the programming language and How that approach can be implemented in Java?
 
William Brogden
Author and all-around good cowpoke
Rancher
Posts: 13074
6
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
In general, you are getting into a huge area of artificial intelligence research applicable to many kinds of pattern recognition, not just acoustic. For simplicity, lets divide the problem.

1. Capture of wave forms of speech samples - the raw data of course is huge so it has to be treated to derive some sort characteristics which can be mathematically manipulated. The FFT - fast fourier transform is popular - you get time series of signal strength in various frequency bins. Of course you can do this in Java.

2. Trying to come up with some mechanism to recognize the characteristics of speakers from the reduced data sets. Personally I have always been attracted to genetic algorithms (thats just one example, a HUGE number of Java experiments with genetic algorithms and other classification schemes can be found.)

Like I said, a huge area - good hunting

Bill
 
Varuna Seneviratna
Ranch Hand
Posts: 170
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
William Brogden wrote:In general, you are getting into a huge area of artificial intelligence research applicable to many kinds of pattern recognition, not just acoustic. For simplicity, lets divide the problem.

1. Capture of wave forms of speech samples - the raw data of course is huge so it has to be treated to derive some sort characteristics which can be mathematically manipulated. The FFT - fast fourier transform is popular - you get time series of signal strength in various frequency bins. Of course you can do this in Java.




How is it to be done using Java?
What is the path to follow?
 
Campbell Ritchie
Sheriff
Pie
Posts: 50251
79
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
What is the algorithm? How would you describe it on paper? Then you can work out what to do in Java™.
 
Bear Bibeault
Author and ninkuma
Marshal
Pie
Posts: 65227
95
IntelliJ IDE Java jQuery Mac Mac OS X
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
It took Dragon Systems over a decade of research and tens of millions of dollars to create Dragon Naturally Speaking. There were over a dozen scientists in the research group, and eight of us in the product group.

I don't think this is a problem you are going to solve on your own from scratch.

Have you googled java speech recognition api?
 
Varuna Seneviratna
Ranch Hand
Posts: 170
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Bear Bibeault wrote:It took Dragon Systems over a decade of research and tens of millions of dollars to create Dragon Naturally Speaking. There were over a dozen scientists in the research group, and eight of us in the product group.

I don't think this is a problem you are going to solve on your own from scratch.

Have you googled java speech recognition api?


The system I have in mind is not as complicated as Dragon Naturally Speaking, what it does is it detects a speaker's identity using the speaker's voice and a word phrase the speaker has previously input to the System Database
 
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic