Digital Signal Processing Reference
In-Depth Information
Chapter 1
Intelligent Audio Analysis: A Definition
Joy, sorrow, tears, lamentation, laughter—to all these music
gives voice, but in such a way that we are transported from the
world of unrest to a world of peace, and see reality in a new
way, as if we were sitting by a mountain lake and contemplating
hills and woods and clouds in the tranquil and fathomless water.
—
Albert Schweitzer
For a start, a short definition of Intelligent Audio Analysis shall be given. This will
be followed by an explanation and clarification of the focus chosen for this topic:
real-life conditions.
1.1 Intelligent Audio Analysis
In general,
audio
is understood as a representation of sound. In this topic, this rep-
resentation is in first given as analogue electrical signal, usually by voltage, then
numerically by digitalisation, i.e., transformation to a pulse-code modulated (PCM)
stream by regularly sampling at uniform intervals in time and quantising to the near-
est value in given digital steps. By that, and in the first place, we deal with mechanical
waves, i.e., a complex series of changes in or oscillation of pressure as compound of
frequencies within the acoustic range available to humans and at sufficiently intense
level to be perceived, i.e.,
audible
by them. These waves may be transmitted by solid,
liquid, or gas—however, in this topic practical examples are limited to air transmis-
sion. This goes, however, without general limitation of the methods presented in
other cases. Further, in this topic, sound is broken down into speech, music, and
general sound. The latter—general sound—may from now on also be referred to as
'sound', omitting 'general' for the sake of simplification.
The
analysis
of audio aims at the extraction of information and—on a higher
level—attachment of semantic meaning to audio signals.
