2 Copyright (C) 2006-2013 Paul Brossier <piem@aubio.org>
4 This file is part of aubio.
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.
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.
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/>.
21 #include "aubio_priv.h"
24 #include "spectral/specdesc.h"
25 #include "spectral/phasevoc.h"
26 #include "spectral/awhitening.h"
27 #include "onset/peakpicker.h"
28 #include "mathutils.h"
29 #include "onset/onset.h"
31 void aubio_onset_default_parameters (aubio_onset_t *o, const char_t * method);
33 /** structure to store object state */
34 struct _aubio_onset_t {
35 aubio_pvoc_t * pv; /**< phase vocoder */
36 aubio_specdesc_t * od; /**< spectral descriptor */
37 aubio_peakpicker_t * pp; /**< peak picker */
38 cvec_t * fftgrain; /**< phase vocoder output */
39 fvec_t * desc; /**< spectral description */
40 smpl_t silence; /**< silence threhsold */
41 uint_t minioi; /**< minimum inter onset interval */
42 uint_t delay; /**< constant delay, in samples, removed from detected onset times */
43 uint_t samplerate; /**< sampling rate of the input signal */
44 uint_t hop_size; /**< number of samples between two runs */
46 uint_t total_frames; /**< total number of frames processed since the beginning */
47 uint_t last_onset; /**< last detected onset location, in frames */
49 uint_t apply_compression;
50 smpl_t lambda_compression;
51 uint_t apply_awhitening; /**< apply adaptive spectral whitening */
52 aubio_spectral_whitening_t *spectral_whitening;
55 /* execute onset detection function on iput buffer */
56 void aubio_onset_do (aubio_onset_t *o, const fvec_t * input, fvec_t * onset)
59 aubio_pvoc_do (o->pv,input, o->fftgrain);
61 if (apply_filtering) {
64 if (o->apply_awhitening) {
65 aubio_spectral_whitening_do(o->spectral_whitening, o->fftgrain);
67 if (o->apply_compression) {
68 cvec_logmag(o->fftgrain, o->lambda_compression);
70 aubio_specdesc_do (o->od, o->fftgrain, o->desc);
71 aubio_peakpicker_do(o->pp, o->desc, onset);
72 isonset = onset->data[0];
74 if (aubio_silence_detection(input, o->silence)==1) {
75 //AUBIO_DBG ("silent onset, not marking as onset\n");
79 uint_t new_onset = o->total_frames + (uint_t)ROUND(isonset * o->hop_size);
80 // check if last onset time was more than minioi ago
81 if (o->last_onset + o->minioi < new_onset) {
82 // start of file: make sure (new_onset - delay) >= 0
83 if (o->last_onset > 0 && o->delay > new_onset) {
86 //AUBIO_DBG ("accepted detection, marking as onset\n");
87 o->last_onset = MAX(o->delay, new_onset);
90 //AUBIO_DBG ("doubled onset, not marking as onset\n");
95 // we are at the beginning of the file
96 if (o->total_frames <= o->delay) {
97 // and we don't find silence
98 if (aubio_silence_detection(input, o->silence) == 0) {
99 uint_t new_onset = o->total_frames;
100 if (o->total_frames == 0 || o->last_onset + o->minioi < new_onset) {
101 isonset = o->delay / o->hop_size;
102 o->last_onset = o->total_frames + o->delay;
107 onset->data[0] = isonset;
108 o->total_frames += o->hop_size;
112 uint_t aubio_onset_get_last (const aubio_onset_t *o)
114 return o->last_onset - o->delay;
117 smpl_t aubio_onset_get_last_s (const aubio_onset_t *o)
119 return aubio_onset_get_last (o) / (smpl_t) (o->samplerate);
122 smpl_t aubio_onset_get_last_ms (const aubio_onset_t *o)
124 return aubio_onset_get_last_s (o) * 1000.;
127 uint_t aubio_onset_set_awhitening (aubio_onset_t *o, uint_t enable)
129 o->apply_awhitening = enable == 1 ? 1 : 0;
133 smpl_t aubio_onset_get_awhitening (aubio_onset_t *o)
135 return o->apply_awhitening;
138 uint_t aubio_onset_set_compression (aubio_onset_t *o, smpl_t lambda)
143 o->lambda_compression = lambda;
144 o->apply_compression = (o->lambda_compression > 0.) ? 1 : 0;
148 smpl_t aubio_onset_get_compression (aubio_onset_t *o)
150 return o->apply_compression ? o->lambda_compression : 0;
153 uint_t aubio_onset_set_silence(aubio_onset_t * o, smpl_t silence) {
154 o->silence = silence;
158 smpl_t aubio_onset_get_silence(const aubio_onset_t * o) {
162 uint_t aubio_onset_set_threshold(aubio_onset_t * o, smpl_t threshold) {
163 aubio_peakpicker_set_threshold(o->pp, threshold);
167 smpl_t aubio_onset_get_threshold(const aubio_onset_t * o) {
168 return aubio_peakpicker_get_threshold(o->pp);
171 uint_t aubio_onset_set_minioi(aubio_onset_t * o, uint_t minioi) {
176 uint_t aubio_onset_get_minioi(const aubio_onset_t * o) {
180 uint_t aubio_onset_set_minioi_s(aubio_onset_t * o, smpl_t minioi) {
181 return aubio_onset_set_minioi (o, (uint_t)ROUND(minioi * o->samplerate));
184 smpl_t aubio_onset_get_minioi_s(const aubio_onset_t * o) {
185 return aubio_onset_get_minioi (o) / (smpl_t) o->samplerate;
188 uint_t aubio_onset_set_minioi_ms(aubio_onset_t * o, smpl_t minioi) {
189 return aubio_onset_set_minioi_s (o, minioi / 1000.);
192 smpl_t aubio_onset_get_minioi_ms(const aubio_onset_t * o) {
193 return aubio_onset_get_minioi_s (o) * 1000.;
196 uint_t aubio_onset_set_delay(aubio_onset_t * o, uint_t delay) {
201 uint_t aubio_onset_get_delay(const aubio_onset_t * o) {
205 uint_t aubio_onset_set_delay_s(aubio_onset_t * o, smpl_t delay) {
206 return aubio_onset_set_delay (o, delay * o->samplerate);
209 smpl_t aubio_onset_get_delay_s(const aubio_onset_t * o) {
210 return aubio_onset_get_delay (o) / (smpl_t) o->samplerate;
213 uint_t aubio_onset_set_delay_ms(aubio_onset_t * o, smpl_t delay) {
214 return aubio_onset_set_delay_s (o, delay / 1000.);
217 smpl_t aubio_onset_get_delay_ms(const aubio_onset_t * o) {
218 return aubio_onset_get_delay_s (o) * 1000.;
221 smpl_t aubio_onset_get_descriptor(const aubio_onset_t * o) {
222 return o->desc->data[0];
225 smpl_t aubio_onset_get_thresholded_descriptor(const aubio_onset_t * o) {
226 fvec_t * thresholded = aubio_peakpicker_get_thresholded_input(o->pp);
227 return thresholded->data[0];
230 /* Allocate memory for an onset detection */
231 aubio_onset_t * new_aubio_onset (const char_t * onset_mode,
232 uint_t buf_size, uint_t hop_size, uint_t samplerate)
234 aubio_onset_t * o = AUBIO_NEW(aubio_onset_t);
236 /* check parameters are valid */
237 if ((sint_t)hop_size < 1) {
238 AUBIO_ERR("onset: got hop_size %d, but can not be < 1\n", hop_size);
240 } else if ((sint_t)buf_size < 2) {
241 AUBIO_ERR("onset: got buffer_size %d, but can not be < 2\n", buf_size);
243 } else if (buf_size < hop_size) {
244 AUBIO_ERR("onset: hop size (%d) is larger than win size (%d)\n", hop_size, buf_size);
246 } else if ((sint_t)samplerate < 1) {
247 AUBIO_ERR("onset: samplerate (%d) can not be < 1\n", samplerate);
251 /* store creation parameters */
252 o->samplerate = samplerate;
253 o->hop_size = hop_size;
255 /* allocate memory */
256 o->pv = new_aubio_pvoc(buf_size, o->hop_size);
257 o->pp = new_aubio_peakpicker();
258 o->od = new_aubio_specdesc(onset_mode,buf_size);
259 o->fftgrain = new_cvec(buf_size);
260 o->desc = new_fvec(1);
261 o->spectral_whitening = new_aubio_spectral_whitening(buf_size, hop_size, samplerate);
263 if (!o->pv || !o->pp || !o->od || !o->fftgrain
264 || !o->desc || !o->spectral_whitening)
267 /* initialize internal variables */
268 aubio_onset_set_default_parameters (o, onset_mode);
270 aubio_onset_reset(o);
278 void aubio_onset_reset (aubio_onset_t *o) {
283 uint_t aubio_onset_set_default_parameters (aubio_onset_t * o, const char_t * onset_mode)
285 uint_t ret = AUBIO_OK;
286 /* set some default parameter */
287 aubio_onset_set_threshold (o, 0.3);
288 aubio_onset_set_delay (o, 4.3 * o->hop_size);
289 aubio_onset_set_minioi_ms (o, 50.);
290 aubio_onset_set_silence (o, -70.);
291 // disable spectral whitening
292 aubio_onset_set_awhitening (o, 0);
293 // disable logarithmic magnitude
294 aubio_onset_set_compression (o, 0.);
296 /* method specific optimisations */
297 if (strcmp (onset_mode, "energy") == 0) {
298 } else if (strcmp (onset_mode, "hfc") == 0 || strcmp (onset_mode, "default") == 0) {
299 aubio_onset_set_threshold (o, 0.058);
300 aubio_onset_set_compression (o, 1.);
301 } else if (strcmp (onset_mode, "complexdomain") == 0
302 || strcmp (onset_mode, "complex") == 0) {
303 aubio_onset_set_delay (o, 4.6 * o->hop_size);
304 aubio_onset_set_threshold (o, 0.15);
305 aubio_onset_set_awhitening(o, 1);
306 aubio_onset_set_compression (o, 1.);
307 } else if (strcmp (onset_mode, "phase") == 0) {
308 o->apply_compression = 0;
309 aubio_onset_set_awhitening (o, 0);
310 } else if (strcmp (onset_mode, "wphase") == 0) {
311 // use defaults for now
312 } else if (strcmp (onset_mode, "mkl") == 0) {
313 aubio_onset_set_threshold (o, 0.05);
314 aubio_onset_set_awhitening(o, 1);
315 aubio_onset_set_compression (o, 0.02);
316 } else if (strcmp (onset_mode, "kl") == 0) {
317 aubio_onset_set_threshold (o, 0.35);
318 aubio_onset_set_awhitening(o, 1);
319 aubio_onset_set_compression (o, 0.02);
320 } else if (strcmp (onset_mode, "specflux") == 0) {
321 aubio_onset_set_threshold (o, 0.18);
322 aubio_onset_set_awhitening(o, 1);
323 aubio_spectral_whitening_set_relax_time(o->spectral_whitening, 100);
324 aubio_spectral_whitening_set_floor(o->spectral_whitening, 1.);
325 aubio_onset_set_compression (o, 10.);
326 } else if (strcmp (onset_mode, "specdiff") == 0) {
327 } else if (strcmp (onset_mode, "old_default") == 0) {
328 // used to reproduce results obtained with the previous version
329 aubio_onset_set_threshold (o, 0.3);
330 aubio_onset_set_minioi_ms (o, 20.);
331 aubio_onset_set_compression (o, 0.);
333 AUBIO_WRN("onset: unknown spectral descriptor type %s, "
334 "using default parameters.\n", onset_mode);
340 void del_aubio_onset (aubio_onset_t *o)
342 if (o->spectral_whitening)
343 del_aubio_spectral_whitening(o->spectral_whitening);
345 del_aubio_specdesc(o->od);
347 del_aubio_peakpicker(o->pp);
349 del_aubio_pvoc(o->pv);
353 del_cvec(o->fftgrain);