C++ Synthesis Tool Kit (STK), MATLAB
- To digitally synthesize the vocal fry (pulse) register, a vocal technique frequently exploited in pop music, and sometimes called “creaky voice.
- To understand the acoustics of the singing voice, particularly its colorful textures, phonations, registers and styles.
- The Liljencrantz-Fant model was adapted to a “best guess” estimate of the derivative glottal excitation waveform that drives the vocal tract during fry. This glottal excitation wave was modeled in MATLAB, and re-formatted for input to the STK objects.
- The STK class VoicForm (which calls SingWave) creates a train of glottal excitations from the input wave.
- Outputs were taken for the 6 phonemes aaa (“hay”), ahh (“wand”), eee (“reed”), ihh (“bit”), ohh (“show”), and ooo (“too”).
- Comparisons were made against self-recorded renditions of vocal fry, and analyzed in MATLAB.
- This short study permitted an investigation into the acoustic and perceptual aspects of vocal fry. However, it was only an introduction to physical modeling of the most advanced musical instrument.
- More realistic sound – the VoicForm object does not sound particularly natural. Hui Ling Lu’s parametric source-filter model sounds much better to the ear.
- Parameterized model with GUI – akin to Perry Cook’s SPASM model, a vocal tract model for users to explore phonations, registers and register hysteresis (transitions).
- A new computer music instrument – this project sparked its inspiration from a popular sample-based vocal synthesizer, Yamaha’s Vocaloid. The Vocaloid is an electronic keyboard of pitches, with additional controls for timbre and phoneme.