Perceptually Motivated Alignment and Interpolation of Pitch-Aligned Time-Frequency Representations

Shahan Nercessian, Jeff Sontag, Alejandro Koretzky

Splice

Accepted to appear at DAFx 2026

Alignment of PTFRs

chroma_src Source chromagram chroma_tgt Target chromagram chroma_prm Aligned chromagram

Method Aligned Source Target Mixdown
None
Argmax
Cross-correlation
Argsort
Argsort + 2-opt
Hungarian
Hungarian + 2-opt

Interpolation of PTFRs

opt_transport_linear Naive Interpolation / Linear OT opt_transport_circular Proposed circular OT along the CoF

t Naive Interpolation / Linear OT Proposed Circular OT + CoF
0
1/3
1/2
2/3
1

Geometric Scale Representation

scale_3d_rep_cylinder Cylindrical scale representation scale_3d_rep_rectangle Rectangular planar profile

Sweep across z-axis (r=0.5)