Information Technology Reference
In-Depth Information
for the contour tests, intervals were allowed to
vary slightly from the comparison phrase, but the
pitch contour remained the same except for one
note. She concluded from her results that musi-
cians identify contour differences more easily
for shorter phrases, while for longer phrases a
changed interval was more easily recognized. She
attributed this to the notion that longer musical
passages afford listeners time to build up a “tonal
framework” which facilitates memory for pitch
intervals. It was not tested whether those with
limited or no musical training establish a similar
frame of reference when listening to music.
Trehub and colleagues (1987) conducted sev-
eral studies with infants as young as six months
to explore their understanding of music. They
discovered that, even at this tender age, infants
have the ability to distinguish between two similar
musical phrases which differ in pitch contour.
In some cases, they can also identify when the
contour remains the same but a pitch interval has
been changed, even if the altered melody was
transposed to a different key from that of the
reference melody.
Smith, Nelson, Grohskopf and Appleton
(1994) performed a study which attempted to
teach pitch interval recognition to two sets of
nonmusical subjects using two different methods.
One group was taught to compare the stimulus
pitch intervals with intervals they already knew
from the beginning notes of three familiar tunes:
Greensleeves
, which begins with a minor third
interval,
Kumbahyah
(a major third), and
Bridal
Chorus
(“Here comes the bride”) (a fourth). The
other group was taught using a more traditional
method from music instruction. It was shown
that the first group performed as well at their
identification task as subjects with musical train-
ing, while the other group performed poorly as
expected.
The work by Smith et al. (1994) also provides
an excellent survey of musical perception stud-
ies as they relate to the nonmusical subject. The
authors concluded this section by saying that
“novices often succeed in musical tasks when
performance has the support of familiar musical
tokens in long-term memory” (p. 46). For MIR
systems designed to be useful to the musically
untrained, this kind of informal music knowledge
common to all could be incorporated in order to
build a system that will be useful to this class
of users.
In discussing some of the difficulties in
obtaining accurate input phrases, McNab et al.
(1996) referred to Sloboda's chapter on music
performance in Deutsch's
The Psychology of
Music
(1982): “Sloboda (1982) reports that people
often distort and recombine melodic fragments
in complex ways, changing melodic contours,
intervals and tonalities; our own studies confirm
this” (p. 14). The resulting challenge to a system
designer is to create music recognition algorithms
which are not overwhelmed computationally in
their efforts to compensate for all of the possible
permutations on a tune based on performance
errors. It is clear that a successful system cannot
rely upon recreating these myriad transformations
from the user's input to the intended song in order
to discover the correct match.
music reproduction
Several studies explored how well musicians and
nonmusicians could manipulate a musical instru-
ment, metronome, or synthesized tone generator in
order to reproduce musical phrases or tempos.
Attneave and Olson (1971) performed experi-
ments to test the ability of musical and of nonmusi-
cal subjects to reproduce note sequences played to
them. This was accomplished by the use of tone
generators under control of the subjects. In the
first experiment subjects were asked to adjust the
output of two tone generators to reproduce two-
note sequences testing interval sizes from 1 to 12
semitones across 8 octaves. In the second experi-
ment, a group of nonmusical subjects performed
a series of trials in which they heard an anchor
tone and manipulated two tone generators so that
Search WWH ::
Custom Search