Memory-Based Approximation of the Gaussian Mixture Model Framework for Bandwidth Extension of Narrowband Speech

Amr Nour-Eldin and Peter Kabal

Multimedia files in this directory

Samples for 2 male and 2 females chosen for their many high-frequency sounds:

sx102_3: Male (DR3/MLNT0)
sx126_1: Female (DR1/FELC0)
sx134_1: Female (DR2/FPAS0)
sx409_1: Male (DR1/MDAB0)

All files use MFCC-based BWE. The "dynamic" version are based on I=128, J=4, Nmin=500, l=3, and Tau=4 (40ms), i.e., narrowband representation has dim 40 (frames at time -120, -80, -40, and 0 ms)

69 kB sx102_3_16kG712FilteredNarrowband.wav
75 kB sx102_3_extended_dynamic.wav
75 kB sx102_3_extended_static.wav
75 kB sx102_3_ms.wav
54 kB sx126_1_16kG712FilteredNarrowband.wav
57 kB sx126_1_extended_dynamic.wav
57 kB sx126_1_extended_static.wav
57 kB sx126_1_ms.wav
64 kB sx134_1_16kG712FilteredNarrowband.wav
66 kB sx134_1_extended_dynamic.wav
65 kB sx134_1_extended_static.wav
67 kB sx134_1_ms.wav
64 kB sx409_1_16kG712FilteredNarrowband.wav
66 kB sx409_1_extended_dynamic.wav
66 kB sx409_1_extended_static.wav
67 kB sx409_1_ms.wav