Reply by Luis Uebel November 30, 20092009-11-30
Try HTK HMM toolkit or Sphinx speech recognizer.
There are some acoustic models and language model for Sphinx in American English.
It is not so robust, but works.
Lu Felipe Uebel

To: s...
From: n...@hotmail.com
Date: Sat, 28 Nov 2009 22:52:51 -0500
Subject: [speech-recognition] Need advices on speech recognition project







Hello, I am an undergraduate student looking to do a senior project on speech recognition. The object of my project is to used speech recognition system to control a device. My partner is working on designing this particular device and my part on this project is the speech recognition. So far on my research, I learned that speech recognition is related to dsp. It uses algorithms to recognized/analysis a speech input and produces a desired output that say it recognized this speech pattern.

What I need advices on is how to go about doing my part. So far from my research, an algorithm called "Hidden Markov Model" is the most efficient way to analysis a speech input. I'm thinking of getting a dsp board to program this algorithm and produce an output in binary code that would go into a microcontroller board that will control a device. What I don't know is how overwhelming programming this algorithm would be. I have experience with basic computer programming and very little C#. I read a dsp overview book before and my partner do have experience in C++. I do know the block diagram for hidden markov model, but do not fully understand it.

My desired output is somehow getting the dsp board to recognized a speech pattern and produce a binary code that represent this pattern so that I can use it to control something. I have 6 months to do this. Is this something doable in 6 months or will programming this algorithms be too much for me and I should focus more on getting some sort of a speech recognition chip and say stick it on a dps board and focus on getting the dsp board to produce the binary code output? Or is there a better way about going at this. If you have any advices, suggestions, or even experience with something like this that can help me out I would really appreciate it.

_____________________________________
Reply by nhan...@hotmail.com November 30, 20092009-11-30
Hello, I am an undergraduate student looking to do a senior project on speech recognition. The object of my project is to used speech recognition system to control a device. My partner is working on designing this particular device and my part on this project is the speech recognition. So far on my research, I learned that speech recognition is related to dsp. It uses algorithms to recognized/analysis a speech input and produces a desired output that say it recognized this speech pattern.

What I need advices on is how to go about doing my part. So far from my research, an algorithm called "Hidden Markov Model" is the most efficient way to analysis a speech input. I'm thinking of getting a dsp board to program this algorithm and produce an output in binary code that would go into a microcontroller board that will control a device. What I don't know is how overwhelming programming this algorithm would be. I have experience with basic computer programming and very little C#. I read a dsp overview book before and my partner do have experience in C++. I do know the block diagram for hidden markov model, but do not fully understand it.

My desired output is somehow getting the dsp board to recognized a speech pattern and produce a binary code that represent this pattern so that I can use it to control something. I have 6 months to do this. Is this something doable in 6 months or will programming this algorithms be too much for me and I should focus more on getting some sort of a speech recognition chip and say stick it on a dps board and focus on getting the dsp board to produce the binary code output? Or is there a better way about going at this. If you have any advices, suggestions, or even experience with something like this that can help me out I would really appreciate it.