cannam@0
|
1
|
cannam@0
|
2 Status of libxtract features implemented via vamp-libxtract.
|
cannam@0
|
3
|
cannam@0
|
4 A feature is described as "plausible" if it runs without crashing on a
|
cannam@0
|
5 test audio file in Sonic Visualiser and produces real valued results
|
cannam@0
|
6 that are not always zero and look (to the uneducated) as if they might
|
cannam@0
|
7 be in a sensible range.
|
cannam@0
|
8
|
cannam@0
|
9 A features is "suspect" if it returns results but they look like
|
cannam@0
|
10 they're all over the place. Other failures are marked accordingly. I
|
cannam@0
|
11 have no idea whether these fail in SV, the plugin, or the extraction
|
cannam@0
|
12 code. The ones that seem to need investigating are marked with a *.
|
cannam@0
|
13
|
cannam@0
|
14 For simple features like mean, variance etc the plugin just extracts
|
cannam@0
|
15 the feature from the spectrum magnitudes of each block.
|
cannam@0
|
16
|
cannam@0
|
17 MEAN Plausible
|
cannam@0
|
18 VARIANCE Plausible
|
cannam@0
|
19 STANDARD_DEVIATION Plausible
|
cannam@0
|
20 AVERAGE_DEVIATION Plausible
|
cannam@0
|
21 SKEWNESS * Suspect
|
cannam@0
|
22 KURTOSIS * Suspect
|
cannam@0
|
23 CENTROID Plausible
|
cannam@0
|
24 IRREGULARITY_K * All zeroes
|
cannam@0
|
25 IRREGULARITY_J * No data
|
cannam@0
|
26 TRISTIMULUS_1 Not implemented (no harmonic spectrum input)
|
cannam@0
|
27 TRISTIMULUS_2 Not implemented (no harmonic spectrum input)
|
cannam@0
|
28 TRISTIMULUS_3 Not implemented (no harmonic spectrum input)
|
cannam@0
|
29 SMOOTHNESS Plausible
|
cannam@0
|
30 SPREAD Plausible
|
cannam@0
|
31 ZCR Plausible
|
cannam@0
|
32 ROLLOFF * Not sure (does produce some data)
|
cannam@0
|
33 LOUDNESS * Does not complete (wild value exposes SV bug?)
|
cannam@0
|
34 FLATNESS Plausible
|
cannam@0
|
35 TONALITY Not implemented (code looks wrong in scalar.c)
|
cannam@0
|
36 CREST Not implemented in libxtract
|
cannam@0
|
37 NOISINESS Not implemented in libxtract
|
cannam@0
|
38 RMS_AMPLITUDE Plausible
|
cannam@0
|
39 INHARMONICITY * Fails to find fundamental frequencies in my test file
|
cannam@0
|
40 POWER Not implemented in libxtract
|
cannam@0
|
41 ODD_EVEN_RATIO Not implemented (no harmonic spectrum input)
|
cannam@0
|
42 SHARPNESS Not implemented in libxtract
|
cannam@0
|
43 SLOPE Not implemented in libxtract
|
cannam@0
|
44 LOWEST_MATCH Not implemented
|
cannam@0
|
45 HPS Not implemented (libxtract code says it doesn't work)
|
cannam@0
|
46 F0 * Fails to find fundamental frequencies in my test file
|
cannam@0
|
47 FLUX Not implemented in libxtract
|
cannam@0
|
48 ATTACK_TIME Not implemented in libxtract
|
cannam@0
|
49 DECAY_TIME Not implemented in libxtract
|
cannam@0
|
50 DELTA_FEATURE Not implemented in libxtract
|
cannam@0
|
51 AUTOCORRELATION Plausible (how should the output be represented?)
|
cannam@0
|
52 AMDF Plausible (how should the output be represented?)
|
cannam@0
|
53 ASDF Plausible (how should the output be represented?)
|
cannam@0
|
54 BARK_COEFFICIENTS Plausible
|
cannam@0
|
55 PEAKS Plausible
|
cannam@0
|
56 MAGNITUDE_SPECTRUM Plausible
|
cannam@0
|
57 AUTOCORRELATION_FFT Plausible (how should the output be represented?)
|
cannam@0
|
58 MFCC * All zeroes
|
cannam@0
|
59 DCT Plausible
|
cannam@0
|
60
|