Active TopicsActive Topics  Display List of Forum MembersMemberlist  CalendarCalendar  Search The ForumSearch  HelpHelp
  RegisterRegister  LoginLogin
French
 FSI Language Courses Forum : Learning Languages : French
Message Icon Topic: Low bit rate = muffled sound Post Reply Post New Topic
Author Message
Waynesail
Newbie
Newbie


Joined: 10 October 2007
Online Status: Offline
Posts: 1
Quote Waynesail Replybullet Topic: Low bit rate = muffled sound
    Posted: 10 October 2007 at 7:31pm
These FSI files are an excellent resource.  However, I noticed that they are a little hard to hear clearly, both on my home computer with no background noise, and even more difficult on my mp3 player.  The sound has a muffled quality.  I see that they are digitized at 32kbps.  It has been my experience that 64kpbs is the minimum necessary (for me, at least) to clearly hear an mp3 file.
 
Is anyone else experiencing any difficulty?
 
Does anyone know if the files are a home digitization of tapes?  Or were they originally created at FSI at only 32 kbps?
 
I'm wondering whether anyone has the files digitized at a higher bit rate that you would be willing to post?
 
Thanks,
 
Waynesail
IP IP Logged
DemiPuppet
Administrator
Administrator


Joined: 27 May 2006
Location: United States
Online Status: Offline
Posts: 163
Quote DemiPuppet Replybullet Posted: 10 October 2007 at 10:13pm
The original source of the audio comes from cassette tapes.  For the audio I supplied I did the following:

1. Digitized the signal by playing the tape with a good Technics cassette player into M-audio FireWire AD converter at 44.1 KHz sample rate. Every 15  tapes or so I would clean and de-magnitize the tape heads.

2. Save the recorded stereo waveform as a 16 bit PCM WAV file.

3. Convert the stereo WAV file into a mono WAV file (sum both channels and divide the amplitude in half). The assumption is that the 1960's era tapes were originally mono. Converting the stereo recording to mono reduces some of the difference noise between the channels.

4. Low pass filter with a sharp 8Khz  corner frequency using a 256 sample window. The highest frequency sibilant in the human voice is about 8Khz.

5. Determine the peak amplitude a and then raise the amplitude of the whole file so that the peak value is 3dB below  the absolute possible peak.  This is done while digitally expanding the signal.  Signals below about -30 db (hiss) will get considerably less amplitude increase than signals above the value.  I did a mild 3dB reduction of any signal less than -50dB.  I didn't do any more than -3dB because I found it too annoying; it would sound like someone is turning the volume up and down.  I notice this on the French Basic audio files (I didn't digitize them).

6. Convert from 44.1KHz down to 22.05KHz  (lossless since the signal was low passed filtered to below 11Khz).

7. Submit the lossless uncompressed WAV files  to this site where they were then converted to  32 kbps MP3s.  I don't know what encoder settings were used.


 
IP IP Logged
Post Reply Post New Topic
Printable version Printable version

Forum Jump
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot delete your posts in this forum
You cannot edit your posts in this forum
You cannot create polls in this forum
You cannot vote in polls in this forum



This page was generated in 0.204 seconds.