1994
The 1994 paper by Kung, Diamantaras, and Taur provides the mathematical engine—the APEX algorithm—that Max Neuhaus later used to bridge the gap between human speech and computer-generated sound in his last major project, Auracle (2004). The APEX algorithm in Auracle: Although the 1994 paper is a technical engineering text, its application in Neuhaus's work transformed it into an artistic tool for "perceptual mapping." Auracle, the system, analyzes a person's voice and extracts a complex set of 43 different acoustic features (pitch, breathiness, jitter, etc.). The APEX algorithm takes this enormous "cloud" of data and reduces it to the three most important (principal) components. "Synthesizer Control": These three extracted components are then used to control a software synthesizer. This allows the user to "play" the instrument using only their voice, with the computer instantly mapping the vocal gestures to complex musical textures. Adaptive Hebbian learning: The paper describes a neural network that "learns" as it goes. For Neuhaus, this was crucial because it allowed the Auracle instrument to adapt to the unique characteristics of different voices, creating a personalized "dialogue" between human and machine.