By
John Robb
The first thing anyone learns about the MP3
audio format is that it is based on a lossy compression algorithm.
Lossy compression means that in order to reduce the file size some
of the original audio has been removed which inherently reduces
the fidelity of the recording. How much data is removed during the
compression process is a based on a combination of factors
including user preference, portability requirements, available
disk space, etc. In the case of an MP3 download the user little
control over how much fidelity has been removed from the audio
source.
In very simple terms the difference in fidelity
can be demonstrated by removing letters from certain words in the
following sentence. This is w_at a sentence wo_ld lo_k like if it
was compres_ed usi_g a_ imaginary los_y compres_ion algorithm. It
doesn’t look very good but I’m sure you can make out the meaning.
That’s lossy compression; enough data is provided in order for you
to understand the meaning even though some of the data is missing.
So how is one to get the fidelity back? The
short answer is you can’t or should I say couldn’t, until now.
There has been a breakthrough in thinking in how to describe audio
wave patterns called the Sound Genome Theory. The Theory states
that digital representation of sound is not just a series of
recorded wave patterns but it is actually made up of “sound
genes”. Similar in concept to DNA a sound spectrogram is made up
of discrete, discernable components that can be “sequenced” and
“synthesized”. This sequence of sound genes is what allows
software based on the Sound Genome Theory to reconstruct the
missing data from a lossy compressed audio file.
Passing our compressed sentence from above
through an imaginary Fidelity Amplifier would give us a sentence
that looks something like this: This is w_at a sentence would
look like if it was compressed usi_g an imaginary lossy
compression algorithm. As you can see increased fidelity
produces a much more pleasing and accurate representation of the
original.
The interesting thing about the Sound Genome
Theory is that it can also be used to compress audio. Software
that is based on the theory can achieve compression comparable or
better than today’s popular lossy compression algorithms yet still
maintain the recorded fidelity. The software can also be used on
existing MP3s in order to achieve incremental lossless data
reduction. Using software based on the Theory will lower the
actual file size or bit rate without loosing fidelity of the
original recording.
Software based on the Sound Genome Theory is
starting to leave the lab now and is available both for playback
of streaming and downloaded audio and as a way to losslessly
compress digital audio. At this point in its evolution the
software is available for use on Windows based PCs. Future
versions of the software are destined for portable devices as well
as other platforms.
Sound Genetics is based out of the high Technology
Triangle from Waterloo Ontario. To start they have
released two products based on the Sound Genome Theory
that are available for download:
Fidelity Amplifier and
Sound Slimmer |
