Kelly,
Here's some stuff audio compression rates as best I understand it.
CD quality digital audio corresponds to a sample rate of 44.1 kHz, on two
channels, encoded at 16 bits which corresponds to a rate of 635 MB/hr.
That's a lot of storage space, especially if you're using flash memory which
is expensive (currently about $1 a MB). Its also takes up a lot of bandwidth
if you want to send it over a network. The same is true of PCM (pulse code
modulation) audio, aka Windows WAV files and Macintosh AIFF files. One can
reduce the sample rate and encode at 8 bits and reduce the number of
channels (i.e. stereo to mono) but with an obvious loss of quality.
The solution is compression schemes (codecs) that use psychoacoutic
principles and other audio features to reduce the bitrate in a ways that
limit the perceived quality loss of the audio stream. Compression formats
include MP3 (MPEG-1 Layer 3), ACC (MPEG-2 AAC), Sony ATRAC, Real Audio RA or
RAM, Microsoft WMA, ADPCM (Adaptive Differential Pulse Code Modulation).
There are many others.
Minidisc, like CD audio, samples at 44.1KHz, in stereo, and encodes in 16
bits but it uses ATRAC (Adaptive Transform Acoustic Coding) to save the
audio in 1/5 the space, e.g. ~127MB/hr without significant loss of quality.
Fraunhofer MP3 (MPEG 1 Layer 3) uses similar principles and saves at a
variety of compression rates. The most common is128kbps, the bitrate made
famous by Napster, which comes out at 58MB/hr. Fraunhofer has more recently
developed a AAC (Advanced Audio Coding) with a number of partners and this
is now considered state of the art, it can encode at a much lower bitrate
and still have the same perceived quality of audio compressed at a much
higher bitrate using other codecs.
Voice recorders use a number of schemes (as do digital telephones and
similar equipment). Voice recording is a much less demanding application as
speech is limited to mid-range frequencies so compression rates can be
greater. Olympus uses DSS (Digital Speech Standard developed by Grundig,
Philips and Olympus) and Sony and Panasonic apparently use variations of
ADPCM. From what I can tell from the online specifications for the more
expensive digital voice recorders when using the highest quality settings
you get:
Olympus DS-2000: 12kHz sample rate; ~6 MB/hr
Panasonic RR-XR320: 16kHz sample rate; ~30 MB/hr
Sony ICD-MS1: 11kHz sample rate; ~15MB/hr
(For comparision: I currently use the Fraunhofer MP3 to store interview
recordings (original made using Minidisc) on my PC at a bitrate of 32kbps or
14MB/hr. This works well for mono recordings (22kHz sample rate with 8 bit
encoding).
The sample rate is significant to the extent it needs to be at least double
the highest frequency one wishes to record. Hence the CD sample rate of
44.1kHz given that the limit of human hearing is about 20kHz.
Voice recorders appear to drop a large range of frequencies, e.g. the
Olympus lists a frequency response of 300 to 5,000 Hz in high quality mode
(compared to 20Hz-20kHz for Minidisc). That's a little better than telephone
quality audio which limits bandwidth to between 400Hz and 3.4kHz but maybe
you don't need to encode much higher frequencies - I wouldn't know at
exactly at what points the frequency response range impacts the audio
quality of speech recordings. There is actually an advantage to limiting
frequency sensitive (e.g. annoying high/low pitched hums from air
conditioners, lighting, etc.).
I'm not sure how the various encoding schemes used by voice recorders
compare. Obvious ATRAC and high bitrate MP3s are vastly superior in the
general sense but the issue is how well voice recorders work for this
application. If they work for you then the technical details probably don't
mean much.
Alan.
|