Marc Bohlen, Jt Rinker: The Universal Whistling Machine
Artist(s):
Title:
- The Universal Whistling Machine
Exhibition:
Creation Year:
- 2004
Medium:
- Machine vision and audio signal processing
Size:
- 11 inches x 8 inches x 9 inches
Category:
Artist Statement:
The Universal Whistling Machine (UWM) is an installation that ponders the phenomenon of whistling as a universal mode of communication, common to digital machines, humans, and many animals. It senses the presence of living, moving creatures in its vicinity and attracts them with a signature whistle. Given a response whistle, UWM counters with its own composition, based on a time-frequency analysis of the original. UWM is an inquiry into automation of an underexposed area of lowbandwidth expression, whistling; direct and immediate; code and content in one. Whistling is admiration, secret code, and protest. Emmet Till, a young man of color, was lynched in 1955 after “wolfwhistling” in the presence of a white woman. Intuitively understood, whistling is transcultural communication below the radar of social etiquette. So much can be expressed by minute alterations of airflow in the mouth. Tongue, throat, lips, and cheeks funnel air into a pressured cocktail of sound energies we use to argue, debate, and sing. But the richness of human language has proven exceedingly difficult to analyze and synthesize, and spoken languages with large vocabularies and multiple speakers still defy the very best speech-recognition systems. A humble machine, UWM deals with a subset of the language problem. Whistling is a communication primitive in most human languages. It is a kind of time travel to a less-articulated state. Inhabitants of Gomera, one of the Canary Islands, use a whistling language, el Silbo Gomera, to communicate from hilltop to hilltop. Their high-power whistles carry farther than the spoken word. We share whistling and song with many animals. Mammals and birds also carry the means for whistling. Just as we carry physical remnants of our bodily evolution in us, we carry the capacity for whistling also. UWM is a statement of affection for the machine, but it is also a critique of the directions we have embarked on. The grand goals of artificial intelligence remain elusive. Maybe we can find peace with machines on more subtle levels. Details are available at: www.buffalo.edu/-mrbohlen/uwm.html
Technical Information:
UWM is programmed in C++ and PD (pure data) under a win32
operating system. It runs on PCs with ieee1394 and audio inputs. People passing by are sensed via a low cost, FireWire-enabled CCD camera/sensor. Sound capture occurs through a noise-reducing microphone array and a standard audio card. Signal sampling occurs at 44.1 kHz. UWM recognizes when people are approaching, invitesthem to whistle, analyzes the response, and reacts in kind when
appropriate. UWM’s whistle synthesizer is based on the basic spectral characteristics of a human whistle. Most human whistles exhibit a fundamental frequency with very few harmonics (often only one or two) as well as a band of high-frequency noise. UWM’s own whistle is created through a process of subtractive synthesis. UWM uses noise as a
signal generator for the whistle synthesizer. The noise is passed through a pair of filters in series. The first is a one-pole, high-pass filter with a roll- off frequency of 600 Hz. The second filter is a bandpass filter, which passes a sinusoid at a specified center frequency and attenuates all other frequencies. The center frequency is the pitch for the whistle, and the “Q” (quality factor or bandwidth) of the
filter is set proportional to the center frequency. Whistle resynthesis and transformation occur in response to input.
The data captured from the pitch tracker are used to mimic or transform the input whistle in order to initiate a dialog with a person. Raw data collected by the pitch tracker is smoothed out by high-threshold gates on pitch and low-threshold gates on amplitude data. UWM is capable of several transformations. An up-transposition or down-transformation is created by adding a fixed pitch interval to the pitch data. This results in a response whistle that is either higher or lower than the input whistle. Contours of the input whistle can also be mapped and increased, decreased or inverted to give a semblance of the whistle with varied pitch transformations. Time transformations are created by making the data-reading rate different than the data-capture rate. This creates responses that are slower or faster but independent of pitch and amplitude. Tempo rubato can be
imitated by randomly changing, within a given range, the time interval between each index of the pitch and amplitude arrays, thus speeding up some portions of the response whistle while slowing down other portions.
