Handle: Engineering Artificial Musical Creativity at the “Trickery” Level - Computational Creativity Research: Towards Creative Machines - page 287

14.5 Three Versions of Handle
14.5.1 Handle as Conductor
Handle cond is both a microcosmic version of the logic-based parts of CAIRA and a
standalone creative T + machine conductor. Can a computing machine “understand”
music and reason from that understanding to the direction of a great conductor, itself
issued in real time so as to improve the performance in question? While we are con-
fident the answer is Yes, the only way to provide via engineering an existence proof
of this affirmative answer is to start humbly, and gradually scale up. Accordingly,
Handle cond was created around a single pianist playing a short piece, and specifically
to investigate the understanding and “conducting” of this playing. A screenshot of
Handle cond in action is shown in Fig. 14.5 .
Handle cond has two major components; the first is an audio analysis module run-
ning withinMATLAB that controls low-level audio and signal processing routines on
incoming live or recorded audio. This module passes information, using a predeter-
mined protocol and format, to the musical calculus system, which runs via Common
Lisp. Handle cond is capable of computing the tempo of live or recorded audio using
the system described in [ 14 ]. The result of this computation is then passed on to
the reasoning system, which in turn determines whether the song is being played
at a tempo appropriate for the audience and context. Figure 14.5 shows Handle cond
responding to a performance of the Prelude in C major from Book 1 of Bach's The
Well-Tempered Clavier by asking for it to be replayed at a slightly faster tempo.
Fig. 14.5 Sample output from Handle cond