Jump to content
Electronics-Lab.com Community

Garvey

Members
  • Posts

    8
  • Joined

  • Last visited

    Never

Posts posted by Garvey

  1. Been looking for this topic for hours.
    Finally, here I am. Well there is a chip or so RSC-364 I found on my PC from last year. It - uses neural network to match on chip speech recognition and accuracy of >97% for speaker independent and 99 for speaker dependent recognition

    www.sensoryinc.com
    Hope we may have solved that part of the project with this.

  2. Thanks for your contribution.
    1. Maybe a VU meter will display the amplitude of the recorded voice say at a predetermined value displayed visually to the user. Such that upon second use - attempt to use voice activation, the user will use the visual VU meter to guage the amplitu de / loudness of his input voice signal. ::)

    2. Another way, may be to trim via software what ever input comes. OR using some f orm of pattern matching (pardon my using the term here) assuming every voice signal has a unique pattern that can be obtained and stored. Thus irrespective of the amplitude -shout or groan the pattern should be CONCEPTUALLY/THEORETICALLY THE SAME ;D

    PLEASE correct me if I am wrong.

  3. Thanks for your contribution.
    I am just back from prevewing the link. That will be a very good addition in
    1. prompting users
    2. And possibly reversing the analogue voice input into text for processing ie working the chip backwards. I don't know it that will be possible.

    Importantly, I have just had a glimpse of Bascom AVR code, looks very simple and appealing in Basic. I hope programming PICs is going to be that simple.

  4. Well using a PC makes the job PC based. And may be some sort of waste cuz I intend to learn by doing as a newbie. Why not use some microcontroller like a DSP or PIC. They act like some RISC microprocessors and I think can do a lot of logic. So you could store pre-recorded voice parterns maybe on another chip for comparison with the incomming voice pattern. Also, human voice should be between 1K and 3K so a Bandstop filter should filter after the microphone.

  5. Yap, Iam a newbie also but I have similar interests in voice and smart card activated/driven systems.

    Now,I contribute;
    Source = Voice - reverb, pressure, frequency range etc
    Recieved by: Sound sensor/ Membrane eg microphonelike device
    Logic - Decrypt sound for the matching sound -in this case Voice - pattern or so
    Action- Based on an algorithm like if (voicepatternrecieved ISEQUALTO John's voicepatternstored ) Then Opendoor
    Next: Some gears/ servos etc move to unlatch the door.

    I consider something more detailed to enable some progress.
    I hope this is useful for now

×
  • Create New...