Possible Issues with voice recognition

Replies

• 

  ice111

  Re: Possible Issues
  
  Voice recognition is quite old, unreliable and not practical. It looks good or exciting in sci-fi movies but in reality it never proved on commercial scale, that's why we are still stuck with more than 100 keys with keyboard, and every day some new keys are added 😀
  
  The major problem it will have is accent problem, "Open" word in voice by an India is different from British to Americans as well as age different too.
  
  Background sound will also be an issue without doubt and accidental voice like open command will also trigger the the mechanism.
  
  I think you better try for some thing else than voice command. With your level of expertise, it could become a headache.
  
  ~ Time
• 

  dgeorgethomas

  Re: Possible Issues
  
  

  ice111

  Voice recognition is quite old, unreliable and not practical. It looks good or exciting in sci-fi movies but in reality it never proved on commercial scale, that's why we are still stuck with more than 100 keys with keyboard, and every day some new keys are added 😀
  
  The major problem it will have is accent problem, "Open" word in voice by an India is different from British to Americans as well as age different too.
  
  Background sound will also be an issue without doubt and accidental voice like open command will also trigger the the mechanism.
  
  I think you better try for some thing else than voice command. With your level of expertise, it could become a headache.
  
  ~ Time

  I thought of this abit and I realize the limitations of voice recognitions are vast. Since in this case the goal would be to open the door I almost think that any sound would be acceptable as long as it was direct communication to the door and not background noise. Therefore would you necessarily need exact recognition of "open"?
• 

  dgeorgethomas

  Google crushed it on there app for the iphone. Very impressive tool to get around the annoying use of a keypad for quick web browsing and 411 look up. I don't think this is a dead technology, only under utilized and misused.
• 

  Ashraf HZ

  You don't need to make it so sophisticated as to recognize the speech to match the word "open".
  
  What you can do is pre-record the command and store it. When the owner wishes to open the door, he/she would recite the same command and the system would correlate it to the stored command.
  
  In terms of noise, its not much of an issue. Background noise can be filtered out, since speech have their own frequency ranges. Just make sure you sample with least 8KHz, and the frame size is appropriate.
  
  Speech recognition has progressed rapidly after the models of the human voice generation was understood, such as the glottal model. But, it still has some quirks to work out, such as distinguishing of homonyms, etc.
  
  Anyway, good luck with the project!
• 

  dgeorgethomas

  ash

  You don't need to make it so sophisticated as to recognize the speech to match the word "open".
  
  What you can do is pre-record the command and store it. When the owner wishes to open the door, he/she would recite the same command and the system would correlate it to the stored command.
  
  In terms of noise, its not much of an issue. Background noise can be filtered out, since speech have their own frequency ranges. Just make sure you sample with least 8KHz, and the frame size is appropriate.
  
  Speech recognition has progressed rapidly after the models of the human voice generation was understood, such as the glottal model. But, it still has some quirks to work out, such as distinguishing of homonyms, etc.
  
  Anyway, good luck with the project!

  Thanks for the reply, i have learned about filters but have not put them into practice so i wasn't sure how effective it would be. I'm surprised the guy above is interested in engineering when he holds those ideals. Science fiction movies and books have been the source of many advancements 😉 Maybe a phasor for my final project lol.
• 

  Ashraf HZ

  Do you have access to Matlab? Thats the best way to test out the speech processing techniques.
  
  Btw, when you say phasor, you mean the phasor gun? 😛
• 

  Predictor

  dgeorgethomas

  I am in third year electronic engineering and we have recently been given a semester long micro processing project. I am currently banging a few ideas around in my head and I'm wondering if I can get some guidance.
  
  The first thought that came to my mind was something voice controlled/voice recognition. I think this is an under used technology and would like to gain more knowledge about it.
  
  My second thought was a modification to handicapped access of doors. The push button system is good and allows for easier access but if you have ever watched a wheel chaired person use it you will notice how difficult it is to orient yourself to the door especially if you were holding something. I was thinking a simple voice controlled door using the "open" command could be very beneficial.
  
  My question is about the possible problems that would occur since I do not have any experience dealing with wave recognition and analysis besides simple filters.
  
  Would background noise be too much of an hindrance to complete this project?
  
  What would other possible complications be, specifically with the audio side of the project?
  
  Do you have any input to possible adaptations using voice recognition?
  
  I'm very new to this forum and any help or input would be very appreciated, basically just to bounce ideas off of someone.

  
  Are you still considering this project? How difficult any of this is, depends on what, exactly, you'd like to accomplish. One issue will be the size of the system's vocabulary. For example, an effective voice recognition system was constructed for understanding the numbers 0 through 9 using a very simple data representation and fuzzy logic. If you just need to recognize the word "open", I don't think that would be too hard.
  
  
  -Will Dwinnell
  Data Mining in MATLAB