Flexible Parametric Voice Coding
Singing
Original (normal)
Reconstruction (no smoothing at all)
Reconstruction (smoothed OQ, T0, AV)
Faster x2
Slower x2
F0 x2
F0 x7/8
OQ x0.5
Original (press)
Reconstruction
Male speech
Original
Reconstruction
Faster x2
Slower x2
F0 x1.5
F0 x2
OQ = 0.2
Female speech
(Lots of false alarms)
Original
Reconstruction
Presentation at CCRMA DSP seminar, Stanford University (4/21/2006)