Lexicon PCM 81 - REV, Руководство пользователя

Страница: 100 / 174

3-39
The Algorithms and Their Parameters
About Pitch Shifting
Digital pitch shifting requires temporal compression or expansion of audio data
and the addition or removal of small audio segments to compensate.
Raising pitch is accomplished by compressing the audio data — making it run
faster. As this shortens the audio segment, a copy of a section of the audio being
processed is added to fill out the remaining time. To lower the pitch, the signal
is expanded, requiring the removal of a section of the audio. These operations
are performed repeatedly at high speed, at intervals which vary according to the
type of audio signal being processed.
At each point where audio is added or removed, the audio signal must be
rejoined, or spliced. If added or removed segments were perfectly correlated
with the remaining audio, splicing would be seamless. In reality, these splice
points are the source of audible artifacts which can be minimized, but not
completely eliminated.
Much of the art of pitch shifting, therefore, is in determining the appropriate
intervals for sampling and the optimal points for splicing. As the PCM 81 Pitch
algorithms gives you significant control over these operations, you may find it
helpful to keep in mind some of the other factors which can affect pitch shift
quality:
• Amount of pitch shifting
Large shift intervals require more splicing. As each splice can be the source of an
artifact, increasing their number increases the likelihood of audible artifacts.
• Pitch-complexity of the input material
Pure tones have simple waveforms which can be correlated easily for smooth
splicing. Pitch shifting of simple tones such as those generated by a solo voice or a
monophonic instrument generally produces fewer artifacts than pitch shifting of wide
bandwidth material such as piano chords, full orchestra, etc.
• Low frequency content of the input material
Pitch shifting low frequency material requires the addition or removal of larger
segments of audio than high frequency material. The increased size of these
segments can introduce delays which are heard as artifacts.
• Transient content of the input material
When large amounts of pitch shifting are applied to complex input material with
dense percussive transients (drums, strummed mandolin, complex speech, etc.)
some of the transients may lose definition or be shifted slightly in time.