/* BeatRoot: An interactive beat tracking system
Copyright (C) 2001, 2006 by Simon Dixon
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program (the file gpl.txt); if not, download it from
http://www.gnu.org/licenses/gpl.txt or write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
package at.ofai.music.beatroot;
import java.util.ListIterator;
import at.ofai.music.util.Event;
import at.ofai.music.util.EventList;
/** Performs tempo induction by finding clusters of similar
* inter-onset intervals (IOIs), ranking them according to the number
* of intervals and relationships between them, and returning a set
* of tempo hypotheses for initialising the beat tracking agents.
*/
class Induction {
/** The maximum difference in IOIs which are in the same cluster */
public static double clusterWidth = 0.025;
/** The minimum IOI for inclusion in a cluster */
public static double minIOI = 0.070;
/** The maximum IOI for inclusion in a cluster */
public static double maxIOI = 2.500;
/** The minimum inter-beat interval (IBI), i.e. the maximum tempo
* hypothesis that can be returned.
* 0.33 seconds == ~181 BPM
* 0.30 seconds == 200 BPM
* 0.25 seconds == 240 BPM
*/
public static double minIBI = 0.33;
/** The maximum inter-beat interval (IBI), i.e. the minimum tempo
* hypothesis that can be returned.
* 1.00 seconds == 60 BPM
* 0.75 seconds == 80 BPM
* 0.60 seconds == 100 BPM
*/
public static double maxIBI = 1.0; // 60BPM // was 0.75 => 80
/** The maximum number of tempo hypotheses to return */
public static int topN = 10;
/** Flag to enable debugging output */
public static boolean debug = false;
/** Performs tempo induction (see JNMR 2001 paper by Simon Dixon for details).
* @param events The onsets (or other events) from which the tempo is induced
* @return A list of beat tracking agents, where each is initialised with one
* of the top tempo hypotheses but no beats
*/
public static AgentList beatInduction(EventList events) {
int i, j, b, bestCount;
boolean submult;
int intervals = 0; // number of interval clusters
int[] bestn = new int[topN];// count of high-scoring clusters
double ratio, err;
int degree;
int maxClusterCount = (int) Math.ceil((maxIOI - minIOI) / clusterWidth);
double[] clusterMean = new double[maxClusterCount];
int[] clusterSize = new int[maxClusterCount];
int[] clusterScore = new int[maxClusterCount];
ListIterator<Event> ptr1, ptr2;
Event e1,e2;
ptr1 = events.listIterator();
while (ptr1.hasNext()) {
e1 = ptr1.next();
ptr2 = events.listIterator();
e2 = ptr2.next();
while (e2 != e1)
e2 = ptr2.next();
while (ptr2.hasNext()) {
e2 = ptr2.next();
double ioi = e2.keyDown - e1.keyDown;
if (ioi < minIOI) // skip short intervals
continue;
if (ioi > maxIOI) // ioi too long
break;
for (b = 0; b < intervals; b++) // assign to nearest cluster
if (Math.abs(clusterMean[b] - ioi) < clusterWidth) {
if ((b < intervals - 1) && (
Math.abs(clusterMean[b+1] - ioi) <
Math.abs(clusterMean[b] - ioi)))
b++; // next cluster is closer
clusterMean[b] = (clusterMean[b] * clusterSize[b] +ioi)/
(clusterSize[b] + 1);
clusterSize[b]++;
break;
}
if (b == intervals) { // no suitable cluster; create new one
if (intervals == maxClusterCount) {
System.err.println("Warning: Too many clusters");
continue; // ignore this IOI
}
intervals++;
for ( ; (b>0) && (clusterMean[b-1] > ioi); b--) {
clusterMean[b] = clusterMean[b-1];
clusterSize[b] = clusterSize[b-1];
}
clusterMean[b] = ioi;
clusterSize[b] = 1;
}
}
}
if (debug) { // output IOI histogram in Matlab format
System.out.println("Inter-onset interval histogram:\n" +
"StartMatlabCode\n" +
"ioi = [");
for (b = 0; b < intervals; b++)
System.out.printf("%4d %7.3f %7d\n",
b, clusterMean[b], clusterSize[b]);
System.out.println("]; ioiclusters(ioi, name);\nEndMatlabCode\n");
}
for (b = 0; b < intervals; b++) // merge similar intervals
// TODO: they are now in order, so don't need the 2nd loop
// TODO: check BOTH sides before averaging or upper gps don't work
for (i = b+1; i < intervals; i++)
if (Math.abs(clusterMean[b] - clusterMean[i]) < clusterWidth) {
clusterMean[b] = (clusterMean[b] * clusterSize[b] +
clusterMean[i] * clusterSize[i]) /
(clusterSize[b] + clusterSize[i]);
clusterSize[b] = clusterSize[b] + clusterSize[i];
--intervals;
for (j = i+1; j <= intervals; j++) {
clusterMean[j-1] = clusterMean[j];
clusterSize[j-1] = clusterSize[j];
}
}
if (intervals == 0)
return new AgentList();
for (b = 0; b < intervals; b++)
clusterScore[b] = 10 * clusterSize[b];
bestn[0] = 0;
bestCount = 1;
for (b = 0; b < intervals; b++)
for (i = 0; i <= bestCount; i++)
if ((i < topN) && ((i == bestCount) ||
(clusterScore[b] > clusterScore[bestn[i]]))){
if (bestCount < topN)
bestCount++;
for (j = bestCount - 1; j > i; j--)
bestn[j] = bestn[j-1];
bestn[i] = b;
break;
}
if (debug) {
System.out.println("Best " + bestCount + " clusters (before):");
for (b = 0; b < bestCount; b++)
System.out.printf("%5.3f : %5d\n", clusterMean[bestn[b]],
clusterScore[bestn[b]]);
}
for (b = 0; b < intervals; b++) // score intervals
for (i = b+1; i < intervals; i++) {
ratio = clusterMean[b] / clusterMean[i];
submult = ratio < 1;
if (submult)
degree = (int) Math.round(1/ratio);
else
degree = (int) Math.round(ratio);
if ((degree >= 2) && (degree <= 8)) {
if (submult)
err = Math.abs(clusterMean[b]*degree - clusterMean[i]);
else
err = Math.abs(clusterMean[b] - clusterMean[i]*degree);
if (err < (submult? clusterWidth : clusterWidth * degree)) {
if (degree >= 5)
degree = 1;
else
degree = 6 - degree;
clusterScore[b] += degree * clusterSize[i];
clusterScore[i] += degree * clusterSize[b];
}
}
}
if (debug) {
System.out.println("Best " + bestCount + " clusters (after):");
for (b = 0; (b < bestCount); b++)
System.out.printf("%5.3f : %5d\n", clusterMean[bestn[b]],
clusterScore[bestn[b]]);
}
if (debug) {
System.out.println("Inter-onset interval histogram 2:");
for (b = 0; b < intervals; b++)
System.out.printf("%3d: %5.3f : %3d (score: %5d)\n",
b, clusterMean[b], clusterSize[b], clusterScore[b]);
}
AgentList a = new AgentList();
for (int index = 0; index < bestCount; index++) {
b = bestn[index];
// Adjust it, using the size of super- and sub-intervals
double newSum = clusterMean[b] * clusterScore[b];
// int newCount = clusterSize[b];
int newWeight = clusterScore[b];
for (i = 0; i < intervals; i++) {
if (i == b)
continue;
ratio = clusterMean[b] / clusterMean[i];
if (ratio < 1) {
degree = (int) Math.round(1 / ratio);
if ((degree >= 2) && (degree <= 8)) {
err = Math.abs(clusterMean[b]*degree - clusterMean[i]);
if (err < clusterWidth) {
newSum += clusterMean[i] / degree * clusterScore[i];
// newCount += clusterSize[i];
newWeight += clusterScore[i];
}
}
} else {
degree = (int) Math.round(ratio);
if ((degree >= 2) && (degree <= 8)) {
err = Math.abs(clusterMean[b] - degree*clusterMean[i]);
if (err < clusterWidth * degree) {
newSum += clusterMean[i] * degree * clusterScore[i];
// newCount += clusterSize[i];
newWeight += clusterScore[i];
}
}
}
}
double beat = newSum / newWeight;
// Scale within range ... hope the grouping isn't ternary :(
while (beat < minIBI) // Maximum speed
beat *= 2.0;
while (beat > maxIBI) // Minimum speed
beat /= 2.0;
if (beat >= minIBI) {
a.add(new Agent(beat));
if (debug)
System.out.printf(" %5.3f", beat);
}
}
if (debug)
System.out.println(" IBI");
return a;
} // beatInduction()
/** For variable cluster widths in newInduction().
* @param low The lowest IOI allowed in the cluster
* @return The highest IOI allowed in the cluster
*/
protected static int top(int low) {
return low + 25; // low/10;
} // top()
/** An alternative (incomplete) tempo induction method (not used).
* Uses integer (millisecond) resolution.
* @param events The events on which tempo induction is performed
*/
public static void newInduction(EventList events) {
final int MAX_MS = 2500;
int[] count = new int[MAX_MS];
for (int i=0; i < MAX_MS; i++)
count[i] = 0;
ListIterator<Event> ptr1, ptr2;
Event e1,e2;
ptr1 = events.listIterator();
while (ptr1.hasNext()) {
e1 = ptr1.next();
ptr2 = events.listIterator();
e2 = ptr2.next();
while (e2 != e1)
e2 = ptr2.next();
while (ptr2.hasNext()) {
e2 = ptr2.next();
int diff = (int) Math.round((e1.keyDown - e2.keyDown) * 1000);
if (diff < MAX_MS)
count[diff]++;
else
break;
}
}
int clnum;
final int MAX_CL = 10;
int[] cluster = new int[MAX_CL];
int[] csize = new int[MAX_CL];
for (clnum = 0; clnum < MAX_CL; clnum++) {
int sum = 0;
int max = 0;
int maxp = 0;
int hi = 70;
int lo = hi;
while (hi < MAX_MS) {
if (hi >= top(lo))
sum -= count[lo++];
else {
sum += count[hi++];
if (sum > max) {
max = sum;
maxp = lo;
}
}
}
if (max == 0)
break;
hi = top(maxp);
if (hi > MAX_MS)
hi = MAX_MS;
int cnt = sum = 0;
for (lo = maxp; lo < hi; lo++) {
sum += lo * count[lo];
cnt += count[lo];
count[lo] = 0;
}
if (cnt != max)
System.err.println("Rounding error in newInduction");
cluster[clnum] = sum / cnt;
csize[clnum] = cnt;
System.out.printf(" %5.3f", sum / 1000.0 / cnt);
//System.out.println("Cluster " + (clnum+1) ": " + (sum/cnt) +
// "ms (" + cnt + " intervals)");
}
System.out.println(" IBI");
// System.out.println("END OF NEW_INDUCTION");
} // newInduction()
} // class Induction