annotate Code/Descriptors/yin/yin.html @ 4:92ca03a8fa99 tip

Update to ICASSP 2013 benchmark
author Dawn Black
date Wed, 13 Feb 2013 11:02:39 +0000
parents
children
rev   line source
Dawn@4 1 <html>
Dawn@4 2 <head>
Dawn@4 3 <title> YIN </title>
Dawn@4 4 </head>
Dawn@4 5
Dawn@4 6 <body>
Dawn@4 7
Dawn@4 8 See README for copyright information.
Dawn@4 9
Dawn@4 10 <hr>
Dawn@4 11 <h2> YIN: fundamental frequency estimator </h2>
Dawn@4 12 <hr>
Dawn@4 13
Dawn@4 14 YIN estimates the fundamental frequency (F0) of an audio signal.
Dawn@4 15 Features are:
Dawn@4 16 <ul>
Dawn@4 17 <li> Reliability (based on tests, see reference below).
Dawn@4 18 <li> Accuracy (subsample resolution).
Dawn@4 19 <li> Wide search range (default is 30 Hz - sr/4).
Dawn@4 20 <li> Good temporal resolution.
Dawn@4 21 <li> Ease of use.
Dawn@4 22 </ul>
Dawn@4 23
Dawn@4 24 YIN operates on vectors or files. YIN outputs a structure containing a set
Dawn@4 25 of four vectors: F0 vs time, two estimates of aperiodic/total power
Dawn@4 26 (one gross estimate, one fine estimate), and a period-smoothed estimate
Dawn@4 27 of instantaneous power.
Dawn@4 28 <p>
Dawn@4 29 If no output argment is specified, YIN
Dawn@4 30 plots F0 as a function of time (in octaves re: 440 Hz), aperiodicity, and power.
Dawn@4 31 <p>
Dawn@4 32 In the F0 plot, samples in blue are reckoned reliable (aperiodicity<threshold),
Dawn@4 33 green are intermediate (aperiodicity<2*threshold), and
Dawn@4 34 yellow unreliable (aperiodicity>2*threshold).
Dawn@4 35
Dawn@4 36 <p>
Dawn@4 37 Type 'help yin' for a description of the parameters. Read the reference below and
Dawn@4 38 the code to understand their meaning. In brief:
Dawn@4 39 <ul>
Dawn@4 40 <li> To increase speed: increase 'hop' or 'minf0'.
Dawn@4 41 <li> To reduce memory needs: reduce 'bufsize', or increase 'hop' or 'minf0'.
Dawn@4 42 <li> To slightly increase reliability: reduce 'hop'.
Dawn@4 43 <li> To slightly increase precision: upsample before processing.
Dawn@4 44 <li> To improve temporal resolution: increase 'minf0', decrease 'hop'.
Dawn@4 45 <li> To process lower F0s: reduce 'minf0'. Higher F0s: upsample and increase 'maxf0'.
Dawn@4 46 <li> To avoid subharmonic errors: increase 'thresh'.
Dawn@4 47 <li> To avoid harmonic/formant errors: reduce 'thresh'.
Dawn@4 48 <li> Make sure that the range [minf0 maxf0] includes the expected f0.
Dawn@4 49 </ul>
Dawn@4 50 Parameter 'thresh' sets the proportion of aperiodic power that is
Dawn@4 51 tolerated within a "periodic" signal. This may vary according to the application.
Dawn@4 52 <p>
Dawn@4 53 For speech or musical instruments a value of 0.1 is usually adequate. Singing voice
Dawn@4 54 may require a smaller value (as low as 0.001) if a harmonic is reinforced by a
Dawn@4 55 sharp formant.
Dawn@4 56 <p>
Dawn@4 57 Some signals are inherently ambiguous. For example
Dawn@4 58 the response of a high-Q resonator excited by a pulse train may be seen either as a complex
Dawn@4 59 tone with an F0 equal to that of the pulse train, or as an
Dawn@4 60 amplitude modulated pure tone with an F0 equal to the resonant frequency.
Dawn@4 61 Neither is more "correct" than the other. To obtain the result
Dawn@4 62 that you expect, you must set the threshold to an appropriate value: small for the
Dawn@4 63 fundamental periodicity, large for the resonance periodicity.
Dawn@4 64
Dawn@4 65 <p>
Dawn@4 66 YIN is described in:
Dawn@4 67 <br></tt>
Dawn@4 68 de Cheveign&eacute;, A., and Kawahara, H. (2002). "YIN, a fundamental frequency estimator
Dawn@4 69 for speech and music," J. Acoust. Soc. Am., 111, 1917-1930. (<a href=
Dawn@4 70 "http://www.ircam.fr/pcm/cheveign/ps/yin.pdf">pdf</a>)
Dawn@4 71 </tt>
Dawn@4 72
Dawn@4 73 <hr>
Dawn@4 74 [Code is <a href="http://www.ircam.fr/pcm/cheveign/sw/yin.zip">here</a>]
Dawn@4 75 [<a href="http://www.ircam.fr/pcm/cheveign">Alain de Cheveign&eacute;</a>]
Dawn@4 76
Dawn@4 77
Dawn@4 78
Dawn@4 79
Dawn@4 80
Dawn@4 81 </body>
Dawn@4 82 </html>
Dawn@4 83
Dawn@4 84