PhaVoRIT: A Phase Vocoder for Real-Time Interactive Time-Stretching
Thorsten Karrer, Eric Lee, Jan Borchers · 2006
Time-based, interactive systems for digital audio and video often employ algorithms to control the speed of the media whilst maintaining the integrity of the content. The phase vocoder is a popular algorithm for time-stretching audio without changing the pitch; its characteristic “transient smearing ” and “reverberation” artifacts, however, limit its application to “simple” audio signals such as instrumental or vocal music. We propose three new methods to further improve the audio quality of phase vocoder-based time-stretching: multiresolution peak-picking accounts for the non-uniform frequency resolution of the human auditory system; sinusoidal trajectory heuristics constrain the set of spectral peaks that are candidates for sinusoidal trajectory continuation, thereby reducing phase propagation along incorrectly detected trajectories; and silent passage phase reset re-establishes the vertical phase coherence at certain intervals. These techniques have been implemented as modules of the PhaVoRIT timestretching software, available as a plug-in to Apple’s Core Audio framework, and the Semantic Time Framework, a multimedia framework we developed for time-based interactive systems. We obtained favorable results when comparing Pha-VoRIT to existing, commonly available audio time-stretching software tools in a formal user study. PhaVoRIT is also deployed as the audio time-stretching engine for Maestro!, an interactive conducting exhibit installed in the Betty Brinn Children’s Museum in Milwaukee, USA. 1