c@277
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
c@277
|
2
|
c@277
|
3 /*
|
c@277
|
4 QM DSP Library
|
c@277
|
5
|
c@277
|
6 Centre for Digital Music, Queen Mary, University of London.
|
c@277
|
7 This file copyright 2008-2009 Matthew Davies and QMUL.
|
c@309
|
8
|
c@309
|
9 This program is free software; you can redistribute it and/or
|
c@309
|
10 modify it under the terms of the GNU General Public License as
|
c@309
|
11 published by the Free Software Foundation; either version 2 of the
|
c@309
|
12 License, or (at your option) any later version. See the file
|
c@309
|
13 COPYING included with this distribution for more information.
|
c@277
|
14 */
|
c@277
|
15
|
c@277
|
16 #include "TempoTrackV2.h"
|
c@277
|
17
|
c@277
|
18 #include <cmath>
|
c@277
|
19 #include <cstdlib>
|
c@278
|
20 #include <iostream>
|
c@277
|
21
|
c@279
|
22 #include "maths/MathUtilities.h"
|
c@277
|
23
|
c@277
|
24 #define EPS 0.0000008 // just some arbitrary small number
|
c@277
|
25
|
c@279
|
26 TempoTrackV2::TempoTrackV2(float rate, size_t increment) :
|
c@279
|
27 m_rate(rate), m_increment(increment) { }
|
c@277
|
28 TempoTrackV2::~TempoTrackV2() { }
|
c@277
|
29
|
c@277
|
30 void
|
c@277
|
31 TempoTrackV2::filter_df(d_vec_t &df)
|
c@277
|
32 {
|
c@278
|
33 d_vec_t a(3);
|
c@278
|
34 d_vec_t b(3);
|
c@278
|
35 d_vec_t lp_df(df.size());
|
c@277
|
36
|
c@278
|
37 //equivalent in matlab to [b,a] = butter(2,0.4);
|
c@278
|
38 a[0] = 1.0000;
|
c@278
|
39 a[1] = -0.3695;
|
c@278
|
40 a[2] = 0.1958;
|
c@278
|
41 b[0] = 0.2066;
|
c@278
|
42 b[1] = 0.4131;
|
c@278
|
43 b[2] = 0.2066;
|
c@278
|
44
|
c@278
|
45 double inp1 = 0.;
|
c@278
|
46 double inp2 = 0.;
|
c@278
|
47 double out1 = 0.;
|
c@278
|
48 double out2 = 0.;
|
c@277
|
49
|
c@277
|
50
|
c@278
|
51 // forwards filtering
|
c@295
|
52 for (unsigned int i = 0;i < df.size();i++)
|
c@278
|
53 {
|
c@278
|
54 lp_df[i] = b[0]*df[i] + b[1]*inp1 + b[2]*inp2 - a[1]*out1 - a[2]*out2;
|
c@278
|
55 inp2 = inp1;
|
c@278
|
56 inp1 = df[i];
|
c@278
|
57 out2 = out1;
|
c@278
|
58 out1 = lp_df[i];
|
c@278
|
59 }
|
c@277
|
60
|
c@278
|
61 // copy forwards filtering to df...
|
c@278
|
62 // but, time-reversed, ready for backwards filtering
|
c@295
|
63 for (unsigned int i = 0;i < df.size();i++)
|
c@278
|
64 {
|
c@278
|
65 df[i] = lp_df[df.size()-i-1];
|
c@278
|
66 }
|
c@277
|
67
|
c@295
|
68 for (unsigned int i = 0;i < df.size();i++)
|
c@278
|
69 {
|
c@278
|
70 lp_df[i] = 0.;
|
c@278
|
71 }
|
c@277
|
72
|
c@278
|
73 inp1 = 0.; inp2 = 0.;
|
c@278
|
74 out1 = 0.; out2 = 0.;
|
c@277
|
75
|
c@277
|
76 // backwards filetering on time-reversed df
|
c@295
|
77 for (unsigned int i = 0;i < df.size();i++)
|
c@278
|
78 {
|
c@278
|
79 lp_df[i] = b[0]*df[i] + b[1]*inp1 + b[2]*inp2 - a[1]*out1 - a[2]*out2;
|
c@278
|
80 inp2 = inp1;
|
c@278
|
81 inp1 = df[i];
|
c@278
|
82 out2 = out1;
|
c@278
|
83 out1 = lp_df[i];
|
c@278
|
84 }
|
c@277
|
85
|
c@277
|
86 // write the re-reversed (i.e. forward) version back to df
|
c@295
|
87 for (unsigned int i = 0;i < df.size();i++)
|
c@278
|
88 {
|
c@278
|
89 df[i] = lp_df[df.size()-i-1];
|
c@278
|
90 }
|
c@277
|
91 }
|
c@277
|
92
|
c@277
|
93
|
c@277
|
94 void
|
c@304
|
95 TempoTrackV2::calculateBeatPeriod(const vector<double> &df,
|
c@304
|
96 vector<double> &beat_period,
|
c@304
|
97 vector<double> &tempi)
|
c@277
|
98 {
|
c@278
|
99 // to follow matlab.. split into 512 sample frames with a 128 hop size
|
c@278
|
100 // calculate the acf,
|
c@278
|
101 // then the rcf.. and then stick the rcfs as columns of a matrix
|
c@278
|
102 // then call viterbi decoding with weight vector and transition matrix
|
c@278
|
103 // and get best path
|
c@277
|
104
|
c@295
|
105 unsigned int wv_len = 128;
|
c@278
|
106 double rayparam = 43.;
|
c@277
|
107
|
c@278
|
108 // make rayleigh weighting curve
|
c@278
|
109 d_vec_t wv(wv_len);
|
c@295
|
110 for (unsigned int i=0; i<wv.size(); i++)
|
c@277
|
111 {
|
c@278
|
112 wv[i] = (static_cast<double> (i) / pow(rayparam,2.)) * exp((-1.*pow(-static_cast<double> (i),2.)) / (2.*pow(rayparam,2.)));
|
c@277
|
113 }
|
c@277
|
114
|
c@278
|
115 // beat tracking frame size (roughly 6 seconds) and hop (1.5 seconds)
|
c@295
|
116 unsigned int winlen = 512;
|
c@295
|
117 unsigned int step = 128;
|
c@278
|
118
|
c@278
|
119 // matrix to store output of comb filter bank, increment column of matrix at each frame
|
c@278
|
120 d_mat_t rcfmat;
|
c@278
|
121 int col_counter = -1;
|
c@278
|
122
|
c@278
|
123 // main loop for beat period calculation
|
c@295
|
124 for (unsigned int i=0; i+winlen<df.size(); i+=step)
|
c@278
|
125 {
|
c@278
|
126 // get dfframe
|
c@278
|
127 d_vec_t dfframe(winlen);
|
c@295
|
128 for (unsigned int k=0; k<winlen; k++)
|
c@278
|
129 {
|
c@278
|
130 dfframe[k] = df[i+k];
|
c@278
|
131 }
|
c@278
|
132 // get rcf vector for current frame
|
c@278
|
133 d_vec_t rcf(wv_len);
|
c@278
|
134 get_rcf(dfframe,wv,rcf);
|
c@277
|
135
|
c@278
|
136 rcfmat.push_back( d_vec_t() ); // adds a new column
|
c@278
|
137 col_counter++;
|
c@295
|
138 for (unsigned int j=0; j<rcf.size(); j++)
|
c@278
|
139 {
|
c@278
|
140 rcfmat[col_counter].push_back( rcf[j] );
|
c@278
|
141 }
|
c@278
|
142 }
|
c@278
|
143
|
c@278
|
144 // now call viterbi decoding function
|
c@278
|
145 viterbi_decode(rcfmat,wv,beat_period,tempi);
|
c@277
|
146 }
|
c@277
|
147
|
c@277
|
148
|
c@277
|
149 void
|
c@277
|
150 TempoTrackV2::get_rcf(const d_vec_t &dfframe_in, const d_vec_t &wv, d_vec_t &rcf)
|
c@277
|
151 {
|
c@278
|
152 // calculate autocorrelation function
|
c@278
|
153 // then rcf
|
c@278
|
154 // just hard code for now... don't really need separate functions to do this
|
c@277
|
155
|
c@278
|
156 // make acf
|
c@277
|
157
|
c@278
|
158 d_vec_t dfframe(dfframe_in);
|
c@277
|
159
|
c@279
|
160 MathUtilities::adaptiveThreshold(dfframe);
|
c@277
|
161
|
c@278
|
162 d_vec_t acf(dfframe.size());
|
c@277
|
163
|
c@278
|
164
|
c@295
|
165 for (unsigned int lag=0; lag<dfframe.size(); lag++)
|
c@278
|
166 {
|
c@278
|
167 double sum = 0.;
|
c@278
|
168 double tmp = 0.;
|
c@277
|
169
|
c@295
|
170 for (unsigned int n=0; n<(dfframe.size()-lag); n++)
|
c@278
|
171 {
|
c@278
|
172 tmp = dfframe[n] * dfframe[n+lag];
|
c@278
|
173 sum += tmp;
|
c@278
|
174 }
|
c@278
|
175 acf[lag] = static_cast<double> (sum/ (dfframe.size()-lag));
|
c@278
|
176 }
|
c@277
|
177
|
c@278
|
178 // now apply comb filtering
|
c@278
|
179 int numelem = 4;
|
c@278
|
180
|
c@295
|
181 for (unsigned int i = 2;i < rcf.size();i++) // max beat period
|
c@278
|
182 {
|
c@278
|
183 for (int a = 1;a <= numelem;a++) // number of comb elements
|
c@278
|
184 {
|
c@278
|
185 for (int b = 1-a;b <= a-1;b++) // general state using normalisation of comb elements
|
c@278
|
186 {
|
c@278
|
187 rcf[i-1] += ( acf[(a*i+b)-1]*wv[i-1] ) / (2.*a-1.); // calculate value for comb filter row
|
c@278
|
188 }
|
c@278
|
189 }
|
c@278
|
190 }
|
c@278
|
191
|
c@278
|
192 // apply adaptive threshold to rcf
|
c@279
|
193 MathUtilities::adaptiveThreshold(rcf);
|
c@278
|
194
|
c@278
|
195 double rcfsum =0.;
|
c@295
|
196 for (unsigned int i=0; i<rcf.size(); i++)
|
c@278
|
197 {
|
c@278
|
198 rcf[i] += EPS ;
|
c@278
|
199 rcfsum += rcf[i];
|
c@278
|
200 }
|
c@277
|
201
|
c@278
|
202 // normalise rcf to sum to unity
|
c@295
|
203 for (unsigned int i=0; i<rcf.size(); i++)
|
c@277
|
204 {
|
c@278
|
205 rcf[i] /= (rcfsum + EPS);
|
c@277
|
206 }
|
c@277
|
207 }
|
c@277
|
208
|
c@277
|
209 void
|
c@278
|
210 TempoTrackV2::viterbi_decode(const d_mat_t &rcfmat, const d_vec_t &wv, d_vec_t &beat_period, d_vec_t &tempi)
|
c@277
|
211 {
|
c@278
|
212 // following Kevin Murphy's Viterbi decoding to get best path of
|
c@278
|
213 // beat periods through rfcmat
|
c@277
|
214
|
c@278
|
215 // make transition matrix
|
c@278
|
216 d_mat_t tmat;
|
c@295
|
217 for (unsigned int i=0;i<wv.size();i++)
|
c@278
|
218 {
|
c@278
|
219 tmat.push_back ( d_vec_t() ); // adds a new column
|
c@295
|
220 for (unsigned int j=0; j<wv.size(); j++)
|
c@278
|
221 {
|
c@278
|
222 tmat[i].push_back(0.); // fill with zeros initially
|
c@278
|
223 }
|
c@278
|
224 }
|
c@278
|
225
|
c@278
|
226 // variance of Gaussians in transition matrix
|
c@278
|
227 // formed of Gaussians on diagonal - implies slow tempo change
|
c@278
|
228 double sigma = 8.;
|
c@278
|
229 // don't want really short beat periods, or really long ones
|
c@295
|
230 for (unsigned int i=20;i <wv.size()-20; i++)
|
c@278
|
231 {
|
c@295
|
232 for (unsigned int j=20; j<wv.size()-20; j++)
|
c@278
|
233 {
|
c@278
|
234 double mu = static_cast<double>(i);
|
c@278
|
235 tmat[i][j] = exp( (-1.*pow((j-mu),2.)) / (2.*pow(sigma,2.)) );
|
c@278
|
236 }
|
c@278
|
237 }
|
c@277
|
238
|
c@278
|
239 // parameters for Viterbi decoding... this part is taken from
|
c@278
|
240 // Murphy's matlab
|
c@277
|
241
|
c@278
|
242 d_mat_t delta;
|
c@278
|
243 i_mat_t psi;
|
c@295
|
244 for (unsigned int i=0;i <rcfmat.size(); i++)
|
c@278
|
245 {
|
c@278
|
246 delta.push_back( d_vec_t());
|
c@278
|
247 psi.push_back( i_vec_t());
|
c@295
|
248 for (unsigned int j=0; j<rcfmat[i].size(); j++)
|
c@278
|
249 {
|
c@278
|
250 delta[i].push_back(0.); // fill with zeros initially
|
c@278
|
251 psi[i].push_back(0); // fill with zeros initially
|
c@278
|
252 }
|
c@278
|
253 }
|
c@277
|
254
|
c@277
|
255
|
c@295
|
256 unsigned int T = delta.size();
|
c@281
|
257
|
c@281
|
258 if (T < 2) return; // can't do anything at all meaningful
|
c@281
|
259
|
c@295
|
260 unsigned int Q = delta[0].size();
|
c@277
|
261
|
c@278
|
262 // initialize first column of delta
|
c@295
|
263 for (unsigned int j=0; j<Q; j++)
|
c@277
|
264 {
|
c@278
|
265 delta[0][j] = wv[j] * rcfmat[0][j];
|
c@278
|
266 psi[0][j] = 0;
|
c@277
|
267 }
|
c@278
|
268
|
c@277
|
269 double deltasum = 0.;
|
c@295
|
270 for (unsigned int i=0; i<Q; i++)
|
c@277
|
271 {
|
c@278
|
272 deltasum += delta[0][i];
|
c@277
|
273 }
|
c@295
|
274 for (unsigned int i=0; i<Q; i++)
|
c@277
|
275 {
|
c@278
|
276 delta[0][i] /= (deltasum + EPS);
|
c@277
|
277 }
|
c@277
|
278
|
c@277
|
279
|
c@295
|
280 for (unsigned int t=1; t<T; t++)
|
c@278
|
281 {
|
c@278
|
282 d_vec_t tmp_vec(Q);
|
c@277
|
283
|
c@295
|
284 for (unsigned int j=0; j<Q; j++)
|
c@278
|
285 {
|
c@295
|
286 for (unsigned int i=0; i<Q; i++)
|
c@278
|
287 {
|
c@278
|
288 tmp_vec[i] = delta[t-1][i] * tmat[j][i];
|
c@278
|
289 }
|
c@278
|
290
|
c@278
|
291 delta[t][j] = get_max_val(tmp_vec);
|
c@277
|
292
|
c@278
|
293 psi[t][j] = get_max_ind(tmp_vec);
|
c@278
|
294
|
c@278
|
295 delta[t][j] *= rcfmat[t][j];
|
c@278
|
296 }
|
c@277
|
297
|
c@278
|
298 // normalise current delta column
|
c@278
|
299 double deltasum = 0.;
|
c@295
|
300 for (unsigned int i=0; i<Q; i++)
|
c@278
|
301 {
|
c@278
|
302 deltasum += delta[t][i];
|
c@278
|
303 }
|
c@295
|
304 for (unsigned int i=0; i<Q; i++)
|
c@278
|
305 {
|
c@278
|
306 delta[t][i] /= (deltasum + EPS);
|
c@278
|
307 }
|
c@278
|
308 }
|
c@277
|
309
|
c@278
|
310 i_vec_t bestpath(T);
|
c@278
|
311 d_vec_t tmp_vec(Q);
|
c@295
|
312 for (unsigned int i=0; i<Q; i++)
|
c@278
|
313 {
|
c@278
|
314 tmp_vec[i] = delta[T-1][i];
|
c@278
|
315 }
|
c@277
|
316
|
c@278
|
317 // find starting point - best beat period for "last" frame
|
c@278
|
318 bestpath[T-1] = get_max_ind(tmp_vec);
|
c@278
|
319
|
c@278
|
320 // backtrace through index of maximum values in psi
|
c@295
|
321 for (unsigned int t=T-2; t>0 ;t--)
|
c@278
|
322 {
|
c@278
|
323 bestpath[t] = psi[t+1][bestpath[t+1]];
|
c@278
|
324 }
|
c@277
|
325
|
c@278
|
326 // weird but necessary hack -- couldn't get above loop to terminate at t >= 0
|
c@278
|
327 bestpath[0] = psi[1][bestpath[1]];
|
c@277
|
328
|
c@295
|
329 unsigned int lastind = 0;
|
c@295
|
330 for (unsigned int i=0; i<T; i++)
|
c@278
|
331 {
|
c@295
|
332 unsigned int step = 128;
|
c@295
|
333 for (unsigned int j=0; j<step; j++)
|
c@278
|
334 {
|
c@278
|
335 lastind = i*step+j;
|
c@278
|
336 beat_period[lastind] = bestpath[i];
|
c@278
|
337 }
|
c@282
|
338 // std::cerr << "bestpath[" << i << "] = " << bestpath[i] << " (used for beat_periods " << i*step << " to " << i*step+step-1 << ")" << std::endl;
|
c@278
|
339 }
|
c@277
|
340
|
c@278
|
341 //fill in the last values...
|
c@295
|
342 for (unsigned int i=lastind; i<beat_period.size(); i++)
|
c@278
|
343 {
|
c@278
|
344 beat_period[i] = beat_period[lastind];
|
c@278
|
345 }
|
c@277
|
346
|
c@295
|
347 for (unsigned int i = 0; i < beat_period.size(); i++)
|
c@277
|
348 {
|
c@279
|
349 tempi.push_back((60. * m_rate / m_increment)/beat_period[i]);
|
c@277
|
350 }
|
c@277
|
351 }
|
c@277
|
352
|
c@277
|
353 double
|
c@277
|
354 TempoTrackV2::get_max_val(const d_vec_t &df)
|
c@277
|
355 {
|
c@278
|
356 double maxval = 0.;
|
c@295
|
357 for (unsigned int i=0; i<df.size(); i++)
|
c@277
|
358 {
|
c@278
|
359 if (maxval < df[i])
|
c@278
|
360 {
|
c@278
|
361 maxval = df[i];
|
c@278
|
362 }
|
c@277
|
363 }
|
c@277
|
364
|
c@278
|
365 return maxval;
|
c@277
|
366 }
|
c@277
|
367
|
c@277
|
368 int
|
c@277
|
369 TempoTrackV2::get_max_ind(const d_vec_t &df)
|
c@277
|
370 {
|
c@278
|
371 double maxval = 0.;
|
c@278
|
372 int ind = 0;
|
c@295
|
373 for (unsigned int i=0; i<df.size(); i++)
|
c@277
|
374 {
|
c@278
|
375 if (maxval < df[i])
|
c@278
|
376 {
|
c@278
|
377 maxval = df[i];
|
c@278
|
378 ind = i;
|
c@278
|
379 }
|
c@277
|
380 }
|
c@278
|
381
|
c@278
|
382 return ind;
|
c@277
|
383 }
|
c@277
|
384
|
c@277
|
385 void
|
c@277
|
386 TempoTrackV2::normalise_vec(d_vec_t &df)
|
c@277
|
387 {
|
c@278
|
388 double sum = 0.;
|
c@295
|
389 for (unsigned int i=0; i<df.size(); i++)
|
c@278
|
390 {
|
c@278
|
391 sum += df[i];
|
c@278
|
392 }
|
c@278
|
393
|
c@295
|
394 for (unsigned int i=0; i<df.size(); i++)
|
c@278
|
395 {
|
c@278
|
396 df[i]/= (sum + EPS);
|
c@278
|
397 }
|
c@277
|
398 }
|
c@277
|
399
|
c@277
|
400 void
|
c@304
|
401 TempoTrackV2::calculateBeats(const vector<double> &df,
|
c@304
|
402 const vector<double> &beat_period,
|
c@304
|
403 vector<double> &beats)
|
c@277
|
404 {
|
c@281
|
405 if (df.empty() || beat_period.empty()) return;
|
c@281
|
406
|
c@278
|
407 d_vec_t cumscore(df.size()); // store cumulative score
|
c@278
|
408 i_vec_t backlink(df.size()); // backlink (stores best beat locations at each time instant)
|
c@278
|
409 d_vec_t localscore(df.size()); // localscore, for now this is the same as the detection function
|
c@277
|
410
|
c@295
|
411 for (unsigned int i=0; i<df.size(); i++)
|
c@277
|
412 {
|
c@278
|
413 localscore[i] = df[i];
|
c@278
|
414 backlink[i] = -1;
|
c@277
|
415 }
|
c@277
|
416
|
c@278
|
417 double tightness = 4.;
|
c@278
|
418 double alpha = 0.9;
|
c@277
|
419
|
c@278
|
420 // main loop
|
c@295
|
421 for (unsigned int i=0; i<localscore.size(); i++)
|
c@278
|
422 {
|
c@278
|
423 int prange_min = -2*beat_period[i];
|
c@278
|
424 int prange_max = round(-0.5*beat_period[i]);
|
c@277
|
425
|
c@278
|
426 // transition range
|
c@278
|
427 d_vec_t txwt (prange_max - prange_min + 1);
|
c@278
|
428 d_vec_t scorecands (txwt.size());
|
c@277
|
429
|
c@295
|
430 for (unsigned int j=0;j<txwt.size();j++)
|
c@278
|
431 {
|
c@278
|
432 double mu = static_cast<double> (beat_period[i]);
|
c@278
|
433 txwt[j] = exp( -0.5*pow(tightness * log((round(2*mu)-j)/mu),2));
|
c@277
|
434
|
c@278
|
435 // IF IN THE ALLOWED RANGE, THEN LOOK AT CUMSCORE[I+PRANGE_MIN+J
|
c@278
|
436 // ELSE LEAVE AT DEFAULT VALUE FROM INITIALISATION: D_VEC_T SCORECANDS (TXWT.SIZE());
|
c@277
|
437
|
c@278
|
438 int cscore_ind = i+prange_min+j;
|
c@278
|
439 if (cscore_ind >= 0)
|
c@278
|
440 {
|
c@278
|
441 scorecands[j] = txwt[j] * cumscore[cscore_ind];
|
c@278
|
442 }
|
c@278
|
443 }
|
c@277
|
444
|
c@278
|
445 // find max value and index of maximum value
|
c@278
|
446 double vv = get_max_val(scorecands);
|
c@278
|
447 int xx = get_max_ind(scorecands);
|
c@277
|
448
|
c@278
|
449 cumscore[i] = alpha*vv + (1.-alpha)*localscore[i];
|
c@278
|
450 backlink[i] = i+prange_min+xx;
|
c@280
|
451
|
c@282
|
452 // std::cerr << "backlink[" << i << "] <= " << backlink[i] << std::endl;
|
c@278
|
453 }
|
c@278
|
454
|
c@278
|
455 // STARTING POINT, I.E. LAST BEAT.. PICK A STRONG POINT IN cumscore VECTOR
|
c@278
|
456 d_vec_t tmp_vec;
|
c@295
|
457 for (unsigned int i=cumscore.size() - beat_period[beat_period.size()-1] ; i<cumscore.size(); i++)
|
c@278
|
458 {
|
c@278
|
459 tmp_vec.push_back(cumscore[i]);
|
c@278
|
460 }
|
c@278
|
461
|
c@278
|
462 int startpoint = get_max_ind(tmp_vec) + cumscore.size() - beat_period[beat_period.size()-1] ;
|
c@278
|
463
|
c@281
|
464 // can happen if no results obtained earlier (e.g. input too short)
|
c@281
|
465 if (startpoint >= backlink.size()) startpoint = backlink.size()-1;
|
c@281
|
466
|
c@278
|
467 // USE BACKLINK TO GET EACH NEW BEAT (TOWARDS THE BEGINNING OF THE FILE)
|
c@278
|
468 // BACKTRACKING FROM THE END TO THE BEGINNING.. MAKING SURE NOT TO GO BEFORE SAMPLE 0
|
c@278
|
469 i_vec_t ibeats;
|
c@278
|
470 ibeats.push_back(startpoint);
|
c@282
|
471 // std::cerr << "startpoint = " << startpoint << std::endl;
|
c@278
|
472 while (backlink[ibeats.back()] > 0)
|
c@278
|
473 {
|
c@282
|
474 // std::cerr << "backlink[" << ibeats.back() << "] = " << backlink[ibeats.back()] << std::endl;
|
c@281
|
475 int b = ibeats.back();
|
c@281
|
476 if (backlink[b] == b) break; // shouldn't happen... haha
|
c@281
|
477 ibeats.push_back(backlink[b]);
|
c@278
|
478 }
|
c@277
|
479
|
c@278
|
480 // REVERSE SEQUENCE OF IBEATS AND STORE AS BEATS
|
c@295
|
481 for (unsigned int i=0; i<ibeats.size(); i++)
|
c@278
|
482 {
|
c@278
|
483 beats.push_back( static_cast<double>(ibeats[ibeats.size()-i-1]) );
|
c@278
|
484 }
|
c@277
|
485 }
|
c@277
|
486
|
c@277
|
487
|