Digital Signal Processing Reference
In-Depth Information
This sequence limits the speed of human speech and thus the speed of information.
Language is a rapid succession of vowels and consonants. The shorter the vowels the less
comprehensible they (and speech generally speaking) become.
There are people who - in order to demonstrate how fast they can think or how intelligent
they are - talk so fast that it is extremely difficult to follow them. It is not, however,
difficult because they think so fast but because our brain has to work very hard to assign
the linguistic elements to individual vowels. Often the whole context is needed if one
wants to get the message. And that is a very stressful business.
Let me now point out the most important results so far: owing to the fact that the human
ear can only hear sinusoidal signals - I will come back to this point later - the following
statement can be made:
Every tone, sound and vowel have some characteristic frequen-
cies which - just like a fingerprint - are virtually unmistakable.
The acoustic recognition of patterns by means of our ears takes place in the frequency
domain because glasses, coins and other solids produce a certain sound when they are
clinked, i.e. they emit - after a transient phase - near- or quasi-periodic signals. The same
is true of vowels. The spectra of these sounds are almost exclusively characterised by only
a few specific frequencies - lines - and thus produce very simple patterns which serve as
an "identity tag".
The acoustic recognition of patterns in nature and in technology
takes place mainly in the frequency domain.
Frequency patterns of near-periodic and quasi-periodic signals -
e.g. vowels - are particularly simple because they merely consist
of several blurred peaks of different height.
Frequency-time landscapes of sounds and vowels resemble
beds
of nails
with nails of different height.
The human acoustic system - ears and brain - functions in very much the same way as
presented and described here in the Illustrations. The human brain does not wait for a
piece of music to end before it begins its frequency analysis but analyses the music on a
continuous basis. Otherwise we would not be able to hear sounds continuously. As I have
already shown in the last chapter, this real-time operation is not carried out through many
successive time windows (windowing) but basically in the ear through many parallel
frequency windows (filters, bandpass filters). The human ear is a FOURIER analyzer, i.e.
a system, which is organized on a frequency basis. Illustration 79 describes the structure
of this chain of filters and its location.
Besides, the human brain must have something like a library or database where the
numerous "beds of nails" are stored as references. How else could we recognize a
particular sound or piece of music?
The acoustic processes in the human ear are much more complex than described here. So
far, it is largely unknown how signals are processed in the brain. It is well-known which
areas of the human brain are responsible for specific functions, but a precise model of
these functions has not been elaborated yet.
