source: src/pitch/pitch.c @ edb1d7a

feature/autosinkfeature/cnnfeature/cnn_orgfeature/constantqfeature/crepefeature/crepe_orgfeature/pitchshiftfeature/pydocstringsfeature/timestretchfix/ffmpeg5pitchshiftsamplertimestretchyinfft+
Last change on this file since edb1d7a was ce3ff2b, checked in by Paul Brossier <piem@piem.org>, 9 years ago

src/pitch/: add const qualifiers, filter_do_outplace to avoid modifying input

  • Property mode set to 100644
File size: 13.9 KB
RevLine 
[96fb8ad]1/*
[e6a78ea]2  Copyright (C) 2003-2009 Paul Brossier <piem@aubio.org>
3
4  This file is part of aubio.
5
6  aubio is free software: you can redistribute it and/or modify
7  it under the terms of the GNU General Public License as published by
8  the Free Software Foundation, either version 3 of the License, or
9  (at your option) any later version.
10
11  aubio is distributed in the hope that it will be useful,
12  but WITHOUT ANY WARRANTY; without even the implied warranty of
13  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  GNU General Public License for more details.
15
16  You should have received a copy of the GNU General Public License
17  along with aubio.  If not, see <http://www.gnu.org/licenses/>.
18
19*/
[96fb8ad]20
21#include "aubio_priv.h"
[6c7d49b]22#include "fvec.h"
23#include "cvec.h"
[a4364b8]24#include "lvec.h"
[96fb8ad]25#include "mathutils.h"
[83963b3]26#include "musicutils.h"
27#include "spectral/phasevoc.h"
[a695854]28#include "temporal/filter.h"
[c159aeb]29#include "temporal/c_weighting.h"
[2d8cffa]30#include "pitch/pitchmcomb.h"
31#include "pitch/pitchyin.h"
32#include "pitch/pitchfcomb.h"
33#include "pitch/pitchschmitt.h"
34#include "pitch/pitchyinfft.h"
[95dc7f2]35#include "pitch/pitchspecacf.h"
[ca1abdd]36#include "pitch/pitch.h"
[96fb8ad]37
[8c3f717]38#define DEFAULT_PITCH_SILENCE -50.
39
[a64ef1d]40/** pitch detection algorithms */
[fddfa64]41typedef enum
42{
[a64ef1d]43  aubio_pitcht_yin,        /**< `yin`, YIN algorithm */
44  aubio_pitcht_mcomb,      /**< `mcomb`, Multi-comb filter */
45  aubio_pitcht_schmitt,    /**< `schmitt`, Schmitt trigger */
46  aubio_pitcht_fcomb,      /**< `fcomb`, Fast comb filter */
47  aubio_pitcht_yinfft,     /**< `yinfft`, Spectral YIN */
[95dc7f2]48  aubio_pitcht_specacf,    /**< `specacf`, Spectral autocorrelation */
[a64ef1d]49  aubio_pitcht_default
50    = aubio_pitcht_yinfft, /**< `default` */
[ca1abdd]51} aubio_pitch_type;
[fe163ad]52
[a64ef1d]53/** pitch detection output modes */
[fddfa64]54typedef enum
55{
[fe163ad]56  aubio_pitchm_freq,   /**< Frequency (Hz) */
57  aubio_pitchm_midi,   /**< MIDI note (0.,127) */
58  aubio_pitchm_cent,   /**< Cent */
59  aubio_pitchm_bin,    /**< Frequency bin (0,bufsize) */
60  aubio_pitchm_default = aubio_pitchm_freq, /**< the one used when "default" is asked */
[ca1abdd]61} aubio_pitch_mode;
[fe163ad]62
[b130600]63/** callback to get pitch candidate, defined below */
[ce3ff2b]64typedef void (*aubio_pitch_detect_t) (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
[6d4ec49]65
[b130600]66/** callback to convert pitch from one unit to another, defined below */
67typedef smpl_t(*aubio_pitch_convert_t) (smpl_t value, uint_t samplerate, uint_t bufsize);
[5284e0d]68
[b130600]69/** callback to fetch the confidence of the algorithm */
70typedef smpl_t (*aubio_pitch_get_conf_t) (void * p);
[f44b111]71
[475da2f]72/** generic pitch detection structure */
[fddfa64]73struct _aubio_pitch_t
74{
[b130600]75  aubio_pitch_type type;          /**< pitch detection mode */
76  aubio_pitch_mode mode;          /**< pitch detection output mode */
77  uint_t samplerate;              /**< samplerate */
[475da2f]78  uint_t bufsize;                 /**< buffer size */
[b130600]79  void *p_object;                 /**< pointer to pitch object */
[fddfa64]80  aubio_filter_t *filter;         /**< filter */
[ce3ff2b]81  fvec_t *filtered;               /**< filtered input */
[fddfa64]82  aubio_pvoc_t *pv;               /**< phase vocoder for mcomb */
83  cvec_t *fftgrain;               /**< spectral frame for mcomb */
84  fvec_t *buf;                    /**< temporary buffer for yin */
[b130600]85  aubio_pitch_detect_t detect_cb; /**< callback to get the pitch candidates */
86  aubio_pitch_convert_t conv_cb;  /**< callback to convert it to the desired unit */
87  aubio_pitch_get_conf_t conf_cb; /**< pointer to the current confidence callback */
[8c3f717]88  smpl_t silence;                 /**< silence threshold */
[96fb8ad]89};
90
[b130600]91/* callback functions for pitch detection */
[ce3ff2b]92static void aubio_pitch_do_mcomb (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
93static void aubio_pitch_do_yin (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
94static void aubio_pitch_do_schmitt (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
95static void aubio_pitch_do_fcomb (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
96static void aubio_pitch_do_yinfft (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
97static void aubio_pitch_do_specacf (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf);
[3ec9d9c]98
[b130600]99/* conversion functions for frequency conversions */
100smpl_t freqconvbin (smpl_t f, uint_t samplerate, uint_t bufsize);
101smpl_t freqconvmidi (smpl_t f, uint_t samplerate, uint_t bufsize);
102smpl_t freqconvpass (smpl_t f, uint_t samplerate, uint_t bufsize);
103
104/* adapter to stack ibuf new samples at the end of buf, and trim `buf` to `bufsize` */
[ce3ff2b]105void aubio_pitch_slideblock (aubio_pitch_t * p, const fvec_t * ibuf);
[3ec9d9c]106
107
[ca1abdd]108aubio_pitch_t *
[ce3ff2b]109new_aubio_pitch (const char_t * pitch_mode,
[168337e]110    uint_t bufsize, uint_t hopsize, uint_t samplerate)
[96fb8ad]111{
[fddfa64]112  aubio_pitch_t *p = AUBIO_NEW (aubio_pitch_t);
[ca1abdd]113  aubio_pitch_type pitch_type;
[fe163ad]114  if (strcmp (pitch_mode, "mcomb") == 0)
[fddfa64]115    pitch_type = aubio_pitcht_mcomb;
[fe163ad]116  else if (strcmp (pitch_mode, "yinfft") == 0)
[97a5878b]117    pitch_type = aubio_pitcht_yinfft;
[fe163ad]118  else if (strcmp (pitch_mode, "yin") == 0)
[fddfa64]119    pitch_type = aubio_pitcht_yin;
[fe163ad]120  else if (strcmp (pitch_mode, "schmitt") == 0)
[fddfa64]121    pitch_type = aubio_pitcht_schmitt;
[fe163ad]122  else if (strcmp (pitch_mode, "fcomb") == 0)
[fddfa64]123    pitch_type = aubio_pitcht_fcomb;
[95dc7f2]124  else if (strcmp (pitch_mode, "specacf") == 0)
125    pitch_type = aubio_pitcht_specacf;
[fe163ad]126  else if (strcmp (pitch_mode, "default") == 0)
[fddfa64]127    pitch_type = aubio_pitcht_default;
[fe163ad]128  else {
[fddfa64]129    AUBIO_ERR ("unknown pitch detection method %s, using default.\n",
130        pitch_mode);
131    pitch_type = aubio_pitcht_default;
[fe163ad]132  }
[2abe563]133
134  // check parameters are valid
135  if ((sint_t)hopsize < 1) {
[d04875c]136    AUBIO_ERR("pitch: got hopsize %d, but can not be < 1\n", hopsize);
[2abe563]137    goto beach;
138  } else if ((sint_t)bufsize < 1) {
[d04875c]139    AUBIO_ERR("pitch: got buffer_size %d, but can not be < 1\n", bufsize);
[2abe563]140    goto beach;
141  } else if (bufsize < hopsize) {
[d04875c]142    AUBIO_ERR("pitch: hop size (%d) is larger than win size (%d)\n", bufsize, hopsize);
[2abe563]143    goto beach;
144  } else if ((sint_t)samplerate < 1) {
[d04875c]145    AUBIO_ERR("pitch: samplerate (%d) can not be < 1\n", samplerate);
[2abe563]146    goto beach;
147  }
148
[b130600]149  p->samplerate = samplerate;
[fe163ad]150  p->type = pitch_type;
[ca1abdd]151  aubio_pitch_set_unit (p, "default");
[6d4ec49]152  p->bufsize = bufsize;
[8c3f717]153  p->silence = DEFAULT_PITCH_SILENCE;
[b130600]154  p->conf_cb = NULL;
[fddfa64]155  switch (p->type) {
[ca1abdd]156    case aubio_pitcht_yin:
[168337e]157      p->buf = new_fvec (bufsize);
[b130600]158      p->p_object = new_aubio_pitchyin (bufsize);
159      p->detect_cb = aubio_pitch_do_yin;
160      p->conf_cb = (aubio_pitch_get_conf_t)aubio_pitchyin_get_confidence;
161      aubio_pitchyin_set_tolerance (p->p_object, 0.15);
[6d4ec49]162      break;
[ca1abdd]163    case aubio_pitcht_mcomb:
[ce3ff2b]164      p->filtered = new_fvec (hopsize);
[168337e]165      p->pv = new_aubio_pvoc (bufsize, hopsize);
166      p->fftgrain = new_cvec (bufsize);
[b130600]167      p->p_object = new_aubio_pitchmcomb (bufsize, hopsize);
[168337e]168      p->filter = new_aubio_filter_c_weighting (samplerate);
[b130600]169      p->detect_cb = aubio_pitch_do_mcomb;
[6d4ec49]170      break;
[ca1abdd]171    case aubio_pitcht_fcomb:
[168337e]172      p->buf = new_fvec (bufsize);
[b130600]173      p->p_object = new_aubio_pitchfcomb (bufsize, hopsize);
174      p->detect_cb = aubio_pitch_do_fcomb;
[6d4ec49]175      break;
[ca1abdd]176    case aubio_pitcht_schmitt:
[168337e]177      p->buf = new_fvec (bufsize);
[b130600]178      p->p_object = new_aubio_pitchschmitt (bufsize);
179      p->detect_cb = aubio_pitch_do_schmitt;
[6d4ec49]180      break;
[ca1abdd]181    case aubio_pitcht_yinfft:
[168337e]182      p->buf = new_fvec (bufsize);
[9c9202f]183      p->p_object = new_aubio_pitchyinfft (samplerate, bufsize);
[b130600]184      p->detect_cb = aubio_pitch_do_yinfft;
185      p->conf_cb = (aubio_pitch_get_conf_t)aubio_pitchyinfft_get_confidence;
186      aubio_pitchyinfft_set_tolerance (p->p_object, 0.85);
[6d4ec49]187      break;
[95dc7f2]188    case aubio_pitcht_specacf:
189      p->buf = new_fvec (bufsize);
190      p->p_object = new_aubio_pitchspecacf (bufsize);
191      p->detect_cb = aubio_pitch_do_specacf;
192      p->conf_cb = (aubio_pitch_get_conf_t)aubio_pitchspecacf_get_tolerance;
193      aubio_pitchspecacf_set_tolerance (p->p_object, 0.85);
194      break;
[6d4ec49]195    default:
196      break;
197  }
198  return p;
[2abe563]199
200beach:
201  AUBIO_FREE(p);
202  return NULL;
[96fb8ad]203}
204
[fddfa64]205void
206del_aubio_pitch (aubio_pitch_t * p)
207{
208  switch (p->type) {
[ca1abdd]209    case aubio_pitcht_yin:
[fddfa64]210      del_fvec (p->buf);
[b130600]211      del_aubio_pitchyin (p->p_object);
[6d4ec49]212      break;
[ca1abdd]213    case aubio_pitcht_mcomb:
[ce3ff2b]214      del_fvec (p->filtered);
[fddfa64]215      del_aubio_pvoc (p->pv);
216      del_cvec (p->fftgrain);
217      del_aubio_filter (p->filter);
[b130600]218      del_aubio_pitchmcomb (p->p_object);
[6d4ec49]219      break;
[ca1abdd]220    case aubio_pitcht_schmitt:
[fddfa64]221      del_fvec (p->buf);
[b130600]222      del_aubio_pitchschmitt (p->p_object);
[6d4ec49]223      break;
[ca1abdd]224    case aubio_pitcht_fcomb:
[fddfa64]225      del_fvec (p->buf);
[b130600]226      del_aubio_pitchfcomb (p->p_object);
[6d4ec49]227      break;
[ca1abdd]228    case aubio_pitcht_yinfft:
[fddfa64]229      del_fvec (p->buf);
[b130600]230      del_aubio_pitchyinfft (p->p_object);
[6d4ec49]231      break;
[95dc7f2]232    case aubio_pitcht_specacf:
233      del_fvec (p->buf);
234      del_aubio_pitchspecacf (p->p_object);
235      break;
[6d4ec49]236    default:
237      break;
238  }
[fddfa64]239  AUBIO_FREE (p);
[96fb8ad]240}
241
[fddfa64]242void
[ce3ff2b]243aubio_pitch_slideblock (aubio_pitch_t * p, const fvec_t * ibuf)
[fddfa64]244{
[6c8ef58]245  uint_t overlap_size = p->buf->length - ibuf->length;
[094a8be]246#if 1 //!HAVE_MEMCPY_HACKS
[6c8ef58]247  uint_t j;
[168337e]248  for (j = 0; j < overlap_size; j++) {
249    p->buf->data[j] = p->buf->data[j + ibuf->length];
[6d4ec49]250  }
[168337e]251  for (j = 0; j < ibuf->length; j++) {
252    p->buf->data[j + overlap_size] = ibuf->data[j];
[6d4ec49]253  }
[6c8ef58]254#else
255  smpl_t *data = p->buf->data;
256  smpl_t *newdata = ibuf->data;
257  memmove(data, data + ibuf->length, overlap_size);
258  memcpy(data + overlap_size, newdata, ibuf->length);
259#endif
[651b97e]260}
261
[fddfa64]262uint_t
[ce3ff2b]263aubio_pitch_set_unit (aubio_pitch_t * p, const char_t * pitch_unit)
[fddfa64]264{
[21e2e6db]265  uint_t err = AUBIO_OK;
[ca1abdd]266  aubio_pitch_mode pitch_mode;
[fe163ad]267  if (strcmp (pitch_unit, "freq") == 0)
[fddfa64]268    pitch_mode = aubio_pitchm_freq;
[53d1497]269  else if (strcmp (pitch_unit, "hertz") == 0)
270    pitch_mode = aubio_pitchm_freq;
[5a2a6c6]271  else if (strcmp (pitch_unit, "Hertz") == 0)
272    pitch_mode = aubio_pitchm_freq;
[53d1497]273  else if (strcmp (pitch_unit, "Hz") == 0)
274    pitch_mode = aubio_pitchm_freq;
275  else if (strcmp (pitch_unit, "f0") == 0)
276    pitch_mode = aubio_pitchm_freq;
[fe163ad]277  else if (strcmp (pitch_unit, "midi") == 0)
[fddfa64]278    pitch_mode = aubio_pitchm_midi;
[fe163ad]279  else if (strcmp (pitch_unit, "cent") == 0)
[fddfa64]280    pitch_mode = aubio_pitchm_cent;
[fe163ad]281  else if (strcmp (pitch_unit, "bin") == 0)
[fddfa64]282    pitch_mode = aubio_pitchm_bin;
[fe163ad]283  else if (strcmp (pitch_unit, "default") == 0)
[fddfa64]284    pitch_mode = aubio_pitchm_default;
[fe163ad]285  else {
[fddfa64]286    AUBIO_ERR ("unknown pitch detection unit %s, using default\n", pitch_unit);
287    pitch_mode = aubio_pitchm_default;
[21e2e6db]288    err = AUBIO_FAIL;
[fe163ad]289  }
290  p->mode = pitch_mode;
[fddfa64]291  switch (p->mode) {
[fe163ad]292    case aubio_pitchm_freq:
[b130600]293      p->conv_cb = freqconvpass;
[fe163ad]294      break;
295    case aubio_pitchm_midi:
[b130600]296      p->conv_cb = freqconvmidi;
[fe163ad]297      break;
298    case aubio_pitchm_cent:
299      /* bug: not implemented */
[b130600]300      p->conv_cb = freqconvmidi;
[fe163ad]301      break;
302    case aubio_pitchm_bin:
[b130600]303      p->conv_cb = freqconvbin;
[fe163ad]304      break;
305    default:
306      break;
307  }
[21e2e6db]308  return err;
[fe163ad]309}
310
[fddfa64]311uint_t
312aubio_pitch_set_tolerance (aubio_pitch_t * p, smpl_t tol)
313{
314  switch (p->type) {
[ca1abdd]315    case aubio_pitcht_yin:
[b130600]316      aubio_pitchyin_set_tolerance (p->p_object, tol);
[7a6cbbe]317      break;
[ca1abdd]318    case aubio_pitcht_yinfft:
[b130600]319      aubio_pitchyinfft_set_tolerance (p->p_object, tol);
[7a6cbbe]320      break;
321    default:
322      break;
323  }
[93177fa]324  return AUBIO_OK;
[f8a38c5]325}
326
[8c3f717]327uint_t
328aubio_pitch_set_silence (aubio_pitch_t * p, smpl_t silence)
329{
[973eb75]330  if (silence <= 0 && silence >= -200) {
[8c3f717]331    p->silence = silence;
332    return AUBIO_OK;
333  } else {
[973eb75]334    AUBIO_ERR("pitch: could not set silence to %.2f", silence);
[8c3f717]335    return AUBIO_FAIL;
336  }
337}
338
339smpl_t
340aubio_pitch_get_silence (aubio_pitch_t * p)
341{
342  return p->silence;
343}
344
[b130600]345
346/* do method, calling the detection callback, then the conversion callback */
[fddfa64]347void
[ce3ff2b]348aubio_pitch_do (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf)
[fddfa64]349{
[b130600]350  p->detect_cb (p, ibuf, obuf);
[8c3f717]351  if (aubio_silence_detection(ibuf, p->silence) == 1) {
352    obuf->data[0] = 0.;
353  }
[b130600]354  obuf->data[0] = p->conv_cb (obuf->data[0], p->samplerate, p->bufsize);
[c078336]355}
356
[b130600]357/* do method for each algorithm */
[fddfa64]358void
[ce3ff2b]359aubio_pitch_do_mcomb (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf)
[fddfa64]360{
[ce3ff2b]361  aubio_filter_do_outplace (p->filter, ibuf, p->filtered);
[fddfa64]362  aubio_pvoc_do (p->pv, ibuf, p->fftgrain);
[b130600]363  aubio_pitchmcomb_do (p->p_object, p->fftgrain, obuf);
364  obuf->data[0] = aubio_bintofreq (obuf->data[0], p->samplerate, p->bufsize);
[c078336]365}
366
[fddfa64]367void
[ce3ff2b]368aubio_pitch_do_yin (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf)
[fddfa64]369{
[6d4ec49]370  smpl_t pitch = 0.;
[fddfa64]371  aubio_pitch_slideblock (p, ibuf);
[b130600]372  aubio_pitchyin_do (p->p_object, p->buf, obuf);
[168337e]373  pitch = obuf->data[0];
374  if (pitch > 0) {
[b130600]375    pitch = p->samplerate / (pitch + 0.);
[168337e]376  } else {
377    pitch = 0.;
[6d4ec49]378  }
[168337e]379  obuf->data[0] = pitch;
[c078336]380}
381
382
[fddfa64]383void
[ce3ff2b]384aubio_pitch_do_yinfft (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * obuf)
[fddfa64]385{
[6d4ec49]386  smpl_t pitch = 0.;
[fddfa64]387  aubio_pitch_slideblock (p, ibuf);
[b130600]388  aubio_pitchyinfft_do (p->p_object, p->buf, obuf);
[168337e]389  pitch = obuf->data[0];
390  if (pitch > 0) {
[b130600]391    pitch = p->samplerate / (pitch + 0.);
[168337e]392  } else {
393    pitch = 0.;
[6d4ec49]394  }
[168337e]395  obuf->data[0] = pitch;
[650e39b]396}
397
[fddfa64]398void
[ce3ff2b]399aubio_pitch_do_specacf (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * out)
[95dc7f2]400{
[c21acb9]401  smpl_t pitch = 0., period;
[95dc7f2]402  aubio_pitch_slideblock (p, ibuf);
403  aubio_pitchspecacf_do (p->p_object, p->buf, out);
404  //out->data[0] = aubio_bintofreq (out->data[0], p->samplerate, p->bufsize);
[c21acb9]405  period = out->data[0];
[95dc7f2]406  if (period > 0) {
407    pitch = p->samplerate / period;
408  } else {
409    pitch = 0.;
410  }
411  out->data[0] = pitch;
412}
413
414void
[ce3ff2b]415aubio_pitch_do_fcomb (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * out)
[fddfa64]416{
417  aubio_pitch_slideblock (p, ibuf);
[b130600]418  aubio_pitchfcomb_do (p->p_object, p->buf, out);
419  out->data[0] = aubio_bintofreq (out->data[0], p->samplerate, p->bufsize);
[c078336]420}
421
[fddfa64]422void
[ce3ff2b]423aubio_pitch_do_schmitt (aubio_pitch_t * p, const fvec_t * ibuf, fvec_t * out)
[fddfa64]424{
[7a6cbbe]425  smpl_t period, pitch = 0.;
[fddfa64]426  aubio_pitch_slideblock (p, ibuf);
[b130600]427  aubio_pitchschmitt_do (p->p_object, p->buf, out);
[168337e]428  period = out->data[0];
429  if (period > 0) {
[b130600]430    pitch = p->samplerate / period;
[168337e]431  } else {
432    pitch = 0.;
[7a6cbbe]433  }
[168337e]434  out->data[0] = pitch;
[96fb8ad]435}
[5284e0d]436
[b130600]437/* conversion callbacks */
438smpl_t
439freqconvbin(smpl_t f, uint_t samplerate, uint_t bufsize)
440{
441  return aubio_freqtobin(f, samplerate, bufsize);
442}
443
444smpl_t
445freqconvmidi (smpl_t f, uint_t samplerate UNUSED, uint_t bufsize UNUSED)
446{
447  return aubio_freqtomidi (f);
448}
449
450smpl_t
451freqconvpass (smpl_t f, uint_t samplerate UNUSED, uint_t bufsize UNUSED)
452{
453  return f;
454}
455
[5284e0d]456/* confidence callbacks */
457smpl_t
458aubio_pitch_get_confidence (aubio_pitch_t * p)
459{
[b130600]460  if (p->conf_cb) {
461    return p->conf_cb(p->p_object);
[5284e0d]462  }
463  return 0.;
464}
Note: See TracBrowser for help on using the repository browser.