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");
78 uint_t new_onset = o->total_frames + (uint_t)ROUND(isonset * o->hop_size);
79 if (o->last_onset + o->minioi < new_onset) {
80 //AUBIO_DBG ("accepted detection, marking as onset\n");
81 o->last_onset = new_onset;
83 //AUBIO_DBG ("doubled onset, not marking as onset\n");
88 // we are at the beginning of the file
89 if (o->total_frames <= o->delay) {
90 // and we don't find silence
91 if (aubio_silence_detection(input, o->silence) == 0) {
92 uint_t new_onset = o->total_frames;
93 if (o->total_frames == 0 || o->last_onset + o->minioi < new_onset) {
94 isonset = o->delay / o->hop_size;
95 o->last_onset = o->total_frames + o->delay;
100 onset->data[0] = isonset;
101 o->total_frames += o->hop_size;
105 uint_t aubio_onset_get_last (const aubio_onset_t *o)
107 return o->last_onset - o->delay;
110 smpl_t aubio_onset_get_last_s (const aubio_onset_t *o)
112 return aubio_onset_get_last (o) / (smpl_t) (o->samplerate);
115 smpl_t aubio_onset_get_last_ms (const aubio_onset_t *o)
117 return aubio_onset_get_last_s (o) * 1000.;
120 uint_t aubio_onset_set_awhitening (aubio_onset_t *o, uint_t enable)
122 o->apply_awhitening = enable == 1 ? 1 : 0;
126 smpl_t aubio_onset_get_awhitening (aubio_onset_t *o)
128 return o->apply_awhitening;
131 uint_t aubio_onset_set_compression (aubio_onset_t *o, smpl_t lambda)
136 o->lambda_compression = lambda;
137 o->apply_compression = (o->lambda_compression > 0.) ? 1 : 0;
141 smpl_t aubio_onset_get_compression (aubio_onset_t *o)
143 return o->apply_compression ? o->lambda_compression : 0;
146 uint_t aubio_onset_set_silence(aubio_onset_t * o, smpl_t silence) {
147 o->silence = silence;
151 smpl_t aubio_onset_get_silence(const aubio_onset_t * o) {
155 uint_t aubio_onset_set_threshold(aubio_onset_t * o, smpl_t threshold) {
156 aubio_peakpicker_set_threshold(o->pp, threshold);
160 smpl_t aubio_onset_get_threshold(const aubio_onset_t * o) {
161 return aubio_peakpicker_get_threshold(o->pp);
164 uint_t aubio_onset_set_minioi(aubio_onset_t * o, uint_t minioi) {
169 uint_t aubio_onset_get_minioi(const aubio_onset_t * o) {
173 uint_t aubio_onset_set_minioi_s(aubio_onset_t * o, smpl_t minioi) {
174 return aubio_onset_set_minioi (o, (uint_t)ROUND(minioi * o->samplerate));
177 smpl_t aubio_onset_get_minioi_s(const aubio_onset_t * o) {
178 return aubio_onset_get_minioi (o) / (smpl_t) o->samplerate;
181 uint_t aubio_onset_set_minioi_ms(aubio_onset_t * o, smpl_t minioi) {
182 return aubio_onset_set_minioi_s (o, minioi / 1000.);
185 smpl_t aubio_onset_get_minioi_ms(const aubio_onset_t * o) {
186 return aubio_onset_get_minioi_s (o) * 1000.;
189 uint_t aubio_onset_set_delay(aubio_onset_t * o, uint_t delay) {
194 uint_t aubio_onset_get_delay(const aubio_onset_t * o) {
198 uint_t aubio_onset_set_delay_s(aubio_onset_t * o, smpl_t delay) {
199 return aubio_onset_set_delay (o, delay * o->samplerate);
202 smpl_t aubio_onset_get_delay_s(const aubio_onset_t * o) {
203 return aubio_onset_get_delay (o) / (smpl_t) o->samplerate;
206 uint_t aubio_onset_set_delay_ms(aubio_onset_t * o, smpl_t delay) {
207 return aubio_onset_set_delay_s (o, delay / 1000.);
210 smpl_t aubio_onset_get_delay_ms(const aubio_onset_t * o) {
211 return aubio_onset_get_delay_s (o) * 1000.;
214 smpl_t aubio_onset_get_descriptor(const aubio_onset_t * o) {
215 return o->desc->data[0];
218 smpl_t aubio_onset_get_thresholded_descriptor(const aubio_onset_t * o) {
219 fvec_t * thresholded = aubio_peakpicker_get_thresholded_input(o->pp);
220 return thresholded->data[0];
223 /* Allocate memory for an onset detection */
224 aubio_onset_t * new_aubio_onset (const char_t * onset_mode,
225 uint_t buf_size, uint_t hop_size, uint_t samplerate)
227 aubio_onset_t * o = AUBIO_NEW(aubio_onset_t);
229 /* check parameters are valid */
230 if ((sint_t)hop_size < 1) {
231 AUBIO_ERR("onset: got hop_size %d, but can not be < 1\n", hop_size);
233 } else if ((sint_t)buf_size < 2) {
234 AUBIO_ERR("onset: got buffer_size %d, but can not be < 2\n", buf_size);
236 } else if (buf_size < hop_size) {
237 AUBIO_ERR("onset: hop size (%d) is larger than win size (%d)\n", hop_size, buf_size);
239 } else if ((sint_t)samplerate < 1) {
240 AUBIO_ERR("onset: samplerate (%d) can not be < 1\n", samplerate);
244 /* store creation parameters */
245 o->samplerate = samplerate;
246 o->hop_size = hop_size;
248 /* allocate memory */
249 o->pv = new_aubio_pvoc(buf_size, o->hop_size);
250 o->pp = new_aubio_peakpicker();
251 o->od = new_aubio_specdesc(onset_mode,buf_size);
252 if (o->od == NULL) goto beach_specdesc;
253 o->fftgrain = new_cvec(buf_size);
254 o->desc = new_fvec(1);
255 o->spectral_whitening = new_aubio_spectral_whitening(buf_size, hop_size, samplerate);
257 /* initialize internal variables */
258 aubio_onset_set_default_parameters (o, onset_mode);
260 aubio_onset_reset(o);
264 del_aubio_peakpicker(o->pp);
265 del_aubio_pvoc(o->pv);
271 void aubio_onset_reset (aubio_onset_t *o) {
276 uint_t aubio_onset_set_default_parameters (aubio_onset_t * o, const char_t * onset_mode)
278 uint_t ret = AUBIO_OK;
279 /* set some default parameter */
280 aubio_onset_set_threshold (o, 0.3);
281 aubio_onset_set_delay (o, 4.3 * o->hop_size);
282 aubio_onset_set_minioi_ms (o, 50.);
283 aubio_onset_set_silence (o, -70.);
284 // disable spectral whitening
285 aubio_onset_set_awhitening (o, 0);
286 // disable logarithmic magnitude
287 aubio_onset_set_compression (o, 0.);
289 /* method specific optimisations */
290 if (strcmp (onset_mode, "energy") == 0) {
291 } else if (strcmp (onset_mode, "hfc") == 0 || strcmp (onset_mode, "default") == 0) {
292 aubio_onset_set_threshold (o, 0.058);
293 aubio_onset_set_compression (o, 1.);
294 } else if (strcmp (onset_mode, "complexdomain") == 0
295 || strcmp (onset_mode, "complex") == 0) {
296 aubio_onset_set_delay (o, 4.6 * o->hop_size);
297 aubio_onset_set_threshold (o, 0.15);
298 aubio_onset_set_awhitening(o, 1);
299 aubio_onset_set_compression (o, 1.);
300 } else if (strcmp (onset_mode, "phase") == 0) {
301 o->apply_compression = 0;
302 aubio_onset_set_awhitening (o, 0);
303 } else if (strcmp (onset_mode, "mkl") == 0) {
304 aubio_onset_set_threshold (o, 0.05);
305 aubio_onset_set_awhitening(o, 1);
306 aubio_onset_set_compression (o, 0.02);
307 } else if (strcmp (onset_mode, "kl") == 0) {
308 aubio_onset_set_threshold (o, 0.35);
309 aubio_onset_set_awhitening(o, 1);
310 aubio_onset_set_compression (o, 0.02);
311 } else if (strcmp (onset_mode, "specflux") == 0) {
312 aubio_onset_set_threshold (o, 0.25);
313 aubio_onset_set_awhitening(o, 1);
314 aubio_onset_set_compression (o, 20.);
315 } else if (strcmp (onset_mode, "specdiff") == 0) {
317 AUBIO_WRN("onset: unknown spectral descriptor type %s, "
318 "using default parameters.\n", onset_mode);
324 void del_aubio_onset (aubio_onset_t *o)
326 del_aubio_spectral_whitening(o->spectral_whitening);
327 del_aubio_specdesc(o->od);
328 del_aubio_peakpicker(o->pp);
329 del_aubio_pvoc(o->pv);
331 del_cvec(o->fftgrain);