Digital Signal Processing Reference
In-Depth Information
Phoneme
is defined as the minimal unit of speech
sound in a language that can serve to dis-
tinguish meaning., 30
Phones
are instantiations of each phoneme category,
represented as individual sounds., 30
Phonetics
concerns the study of speech sounds and
their physical properties, their articulatory
and acoustic properties., 30
Phonology
refers to the characterization of speech
sounds as abstract units, together with their
realization rules depending on the phonetic
context., 30
Phonotactics
of a language are constraints in the way vow-
els and consonants are combined into sylla-
bles., 32
Temporal downsampling
corresponds to decrease the frame rate at the
input of a second level hierarchical scheme,
reducing computational time., 44
Window downsampling
is a technique that removes posterior vec-
tors contained in the input window of a sec-
ond level hierarchical scheme. These vectors
contain irrelevant information for the task of
phoneme recognition., 46