Movie demonstration from conference
Here are several examples discussed in our article B.L. Sturm, C.Roads, A.McLeran, and J.J. Shynk. "Analysis, visualization, and transformation of audio signals using overcomplete methods." In Proc. Int. Computer Music Conf., Belfast, Ireland, Aug. 2008 (submitted).
|
Original Signals |
||
| Bach Organ | ||
| Funk Drums | ||
| Jazz Drums | ||
| Flute 1 | ||
| Flute 2 | ||
| Glockenspiel1 | ||
| Harpsichord | ||
| Saxophone | ||
| Industrial | ||
| French Speech | ||
| Processing | Result | Description |
| Piano | Piano | Low order approximation (below 100 atoms), multiplies frequencies by 4 with a bleed factor of 8 |
| Piano | Multiplies frequencies by 0.5 with a bleed factor of 2 | |
| Piano | Multiplies frequencies by 2 and with a bleed factor of 4 | |
| Piano | Multiplies atom positions and durations by 1.5 | |
| Piano | Replaces gaussian window with a fof window | |
| Piano | Adjusts atom frequencies to be the nearest multiple of 40 hz | |
| Glockenspiel2 | Glockenspiel2 | Atoms with bleed factor of 3 |
| Glockenspiel2 | Only includes atoms of low order (first 1000 atoms) with bleed factor of 3 | |
| Glockenspiel2 | If atoms are longer than 5 ms, then they have a bleed factor of 8 | |
| Glockenspiel2 | If atoms are longer than 10 ms, then they have a bleed factor of 8 and are an octave down | |
| Atom Bleed | Funk Drums | Bleed factor of 2 |
| Funk Drums | Bleed factor of 5 | |
| Saxophone | Bleed factor of 5 | |
| Flute 2 | Bleed factor of 5 | |
| Flute 1 | Bleed factor of 0.01 (shrink) | |
| Flute 1 | Bleed factor of 0.1 (shrink) | |
| Coalescence | Bach | Gradually build up the sound |
| Flute 2 | Random increase in reconstruction density and atom duration | |
| Disintegration | Industrial | Gradually decay the sound |
| Frequency Filtering | Flute 1 | Pass atoms with modulation frequencies less than 1 kHz |
| Saxophone | Pass atoms with modulation frequencies less than 500 Hz | |
| Jazz Drums | Pass atoms with modulation frequencies above 1.5 kHz | |
| Saxophone | Pass atoms with modulation frequencies above 1.5 kHz | |
| Energy Filtering | Funk Drums | Keep low energy atoms |
| Funk Drums | Keep atoms in top 30 dB | |
| Glockenspiel | Remove atoms in top 48 dB | |
| Harpsichord | Keep atoms with energies between 30-36 dB | |
| Scale Filtering | Glockenspiel | Keep atoms with scales equal to or larger than 2048 samples |
| Glockenspiel | Keep atoms with scales less than 2048 samples | |
| Speech | Keep atoms with scales equal to 256 samples | |
| Naive Pitch Shift | Flute 1 | Transposed down two octaves |
| Glockenspiel | Transposed down one octave | |
| Examples from Audio and Music Computing for Multimedia Workshop, ACM Multimedia Conference 2006 | ||
| Harmonization | Flute 1 | Modulate atoms one octave and add to original signal |
| Saxophone | Transposed down one octave and up one octave and then mixed with original book | |
| Jazz Drums | Transposed down one octave and up one octave and then mixed with original book | |
| Flute 2 | Transposed down one octave and up one octave and then mixed with original book | |
| Jitter | Funk Drums | Shift atom onsets randomly by up to 50% of scale |
| Flute 2 | Shift atom onsets randomly by up to 20% of total number of samples | |
| Jazz Drums | Shift atom onsets randomly by up to 20% of total number of samples | |
| Flute 2 | Shift atom onsets randomly by up to 20% of nyquist frequency | |
| Saxophone | Shift frequencies randomly by up to 20% of nyquist frequency | |
| Spatialize | Harpsichord | Start in center, shift larger scale atoms to right, shorter scale atoms to left, then return to center |
| Substitution | Flute 1 | Analyzed with Gabor, synthesized with damped sinusoids |
| Speech | Analyzed with Gabor, synthesized with damped sinusoids, mixed with original | |
| Naive Time Scale | Funk Drums | Dilated 200% |
| Flute 1 | Dilated 75% | |
| Jazz Drums | Scaled duration and time onsets by 400% | |
| Saxophone | Scaled duration and time onsets by 400% | |
| Flute 2 | Scaled duration and time onsets by 400% | |