Vowel-Based Multi-Channel Decorrelation

A novel audio decorrelation method that integrates velvet noise with parametric modelling of vowel filters derived from recorded speech. By capturing vocal timbre through Linear Predictive Coding (LPC), the technique enables the creation of spatially diffuse, voice-inspired multi-channel reverberation — including 5.1 reverberation from a mono input — with personalisation through the user’s own recorded voice.

A collaboration with Michele Pizzi (independent researcher, spatial audio DSP). I contributed psychoacoustic evaluation methodology and listening test infrastructure at the Department of Multimedia Systems recording studio. Published in the International Journal of Electronics and Telecommunications (2026). DOI: 10.24425/ijet.2026.157864

@article{pizzi_using_2026,
  author  = {Pizzi, Michele and Mróz, Bartłomiej},
  title   = {Using {Vowel} {Characteristics} for {Multi}-channel {Signal} {Decorrelation} and {Reverberation}},
  journal = {International Journal of Electronics and Telecommunications},
  year    = {2026},
  pages   = {1--9},
  doi     = {10.24425/ijet.2026.157864}
}