... but then I read that even at 128kbps, there is
It would be helpful to know for the DMR-300 (and other
potential MP3/voice recorders) whether they can sample
a 11khz signal of the human voice, say at 16 bits per
sample, and produce a distortion-free signal. The
arithmetic (11000x16=176kbps) seems to indicate that
this is possible, since it shouldn't be difficult to
compress this losslessly to 128kbps or even 64kbps.
Note that it's critical to be able to control the
sampling rate ... the top voice recognition
program, Dragon NaturallySpeaking samples at 11khz,
and I believe that the #2 program, ViaVoice samples at
22khz. There IS NO WAY to take an 8khz sample, or
even a 16khz sample, and turn it into an 11 khz signal
suitable for voice recognition. Note that it's the
sampling rate that matters, not the bit rate.
I'm ranting here only because I am unable to find any
technical information about the DMR-300 online, other
than Tom's review and the CNET review:
The "knowbrainer" you link to is a VAR Dragon Voice recognition dealer & really knows his stuff.
Mic placement & consistant voice pattern is more important than frequency response of mic..
a $49. headset mic can get excellant results..The best VR mics are less than $100.
An acomplished Dragon user can get better results with a $9. headset than a novice can with a $100. mic.
proper software 'training', precise mic placement, careful cadence & diction = good results.
pointer: when a 'voice recognition mistake' is found train & retrain & retrain THAT WORD.. if you correct each mistake with voice recognition & not the keyboard the software will learn YOUR accent / speech pattern.
& always backup your 'training files'.. a crash or corrupted file can mean loosing 100s of hours of training.