2 aubiopitch - a command line tool to extract musical pitch
7 aubiopitch [[-i] source] [-o sink]
8 [-r rate] [-B win] [-H hop]
9 [-p method] [-u unit] [-l thres]
16 aubiopitch attempts to detect the pitch, the perceived height of a musical
19 When started with an input source (-i/--input), the detected pitch are
20 printed on the console, prefixed by a timestamp in seconds. If no pitch
21 candidate is found, the output is 0.
23 When started without an input source, or with the jack option (-j/--jack),
24 aubiopitch starts in jack mode.
28 This program follows the usual GNU command line syntax, with long options
29 starting with two dashes (--). A summary of options is included below.
31 -i, --input source Run analysis on this audio file. Most uncompressed and
32 compressed are supported, depending on how aubio was built.
34 -o, --output sink Save results in this file. The file will be created on
35 the model of the input file. The detected frequency is played at the
38 -r, --samplerate rate Fetch the input source, resampled at the given
39 sampling rate. The rate should be specified in Hertz as an integer. If 0,
40 the sampling rate of the original source will be used. Defaults to 0.
42 -B, --bufsize win The size of the buffer to analyze, that is the length
43 of the window used for spectral and temporal computations. Defaults to 2048.
45 -H, --hopsize hop The number of samples between two consecutive analysis.
48 -p, --pitch method The pitch detection method to use. See PITCH METHODS
49 below. Defaults to 'default'.
51 -u, --pitch-unit unit The unit to be used to print frequencies. Possible
52 values include midi, bin, cent, and Hz. Defaults to 'Hz'.
54 -l, --pitch-tolerance thres Set the tolerance for the pitch detection
55 algorithm. Typical values range between 0.2 and 0.9. Pitch candidates found
56 with a confidence less than this threshold will not be selected. The higher
57 the threshold, the more confidence in the candidates. Defaults to unset.
59 -s, --silence sil Set the silence threshold, in dB, under which the onset
60 will not be detected. A value of -20.0 would eliminate most onsets but the
61 loudest ones. A value of -90.0 would select all onsets. Defaults to -90.0.
63 -T, --timeformat format Set time format (samples, ms, seconds). Defaults to
66 -m, --mix-input Mix source signal to the output signal before writing to
69 -f, --force-overwrite Overwrite output file if it already exists.
71 -j, --jack Use Jack input/output. You will need a Jack connection
72 controller to feed aubio some signal and listen to its output.
74 -h, --help Print a short help message and exit.
76 -v, --verbose Be verbose.
80 Available methods are:
82 default use the default method
84 Currently, the default method is set to yinfft.
86 schmitt Schmitt trigger
88 This pitch extraction method implements a Schmitt trigger to estimate the
89 period of a signal. It is computationally very inexpensive, but also very
92 fcomb a fast harmonic comb filter
94 This pitch extraction method implements a fast harmonic comb filter to
95 determine the fundamental frequency of a harmonic sound.
97 mcomb multiple-comb filter
99 This fundamental frequency estimation algorithm implements spectral
100 flattening, multi-comb filtering and peak histogramming.
102 specacf Spectral auto-correlation function
106 This algorithm was developed by A. de Cheveigne and H. Kawahara and
107 was first published in:
109 De Cheveigné, A., Kawahara, H. (2002) "YIN, a fundamental frequency
110 estimator for speech and music", J. Acoust. Soc. Am. 111, 1917-1930.
112 yinfft Yinfft algorithm
114 This algorithm was derived from the YIN algorithm. In this implementation, a
115 Fourier transform is used to compute a tapered square difference function,
116 which allows spectral weighting. Because the difference function is tapered,
117 the selection of the period is simplified.
119 Paul Brossier, Automatic annotation of musical audio for interactive systems,
120 Chapter 3, Pitch Analysis, PhD thesis, Centre for Digital music, Queen Mary
121 University of London, London, UK, 2006.
123 yinfast YIN algorithm (accelerated)
125 An optimised implementation of the YIN algorithm, yielding results identical
126 to the original YIN algorithm, while reducing its computational cost from
127 O(n^2) to O(n log(n)).
141 This manual page was written by Paul Brossier <piem@aubio.org>. Permission is
142 granted to copy, distribute and/or modify this document under the terms of
143 the GNU General Public License as published by the Free Software Foundation,
144 either version 3 of the License, or (at your option) any later version.