Computational Biology and Language - Ambient Intelligence for Scientific Discovery - page 34

Fig. 6. Digital Representation. (A) Digital waveform of a speech signal of the utterance
“how to recognize speech with this new display”. The x-axis shows the time, while the
y-axis shows the amplitude (loudness) of the signal. (B) Protein signal, where the
sequence is represented by a numerical scale of charge of the residues. (C) Protein
signal in terms of hydrophobicity of the residues. X-axis in (B) and (C) shows the
residue number along the length of the protein, and y-axis is the value of the property
of the residues (here charge or hydrophobicity).

a word in speech amounts to recognizing these phones. There are typically 50
phones in speech. Thus, identification of “word” equivalents in protein sequences
using the signal processing approach is equivalent to “phone” identification.

3.2 Information Required to Decode the Signal

The content of a speech signal is not only dependent on the signal itself; its
interpretation relies on an external entity, the listener. For example, consider
the phrases:

How to recognize speech with this new display
How to wreck a nice beach with this nudist play

The two phrases are composed of almost identical phone sequences, but result in
two different sentences. Spectrograms showing the frequency decomposition of
the sound signals are shown in Fig. 7B and 7C, for these two sentences spoken
by the same speaker.
Given the speech signal or a spectrogram, which utterance was meant by the
speaker can be found by the context in which it was spoken. Thus the complete
information for interpretation is not contained in the speech signal itself, but is
inferred from the context. On the other hand, the linear strings of amino acids
that make up a protein contain in principle all the information needed to fold it
into a 3-D shape capable of fulfilling its designated function.