/**
* Portions Copyright 2006 DFKI GmbH.
* Portions Copyright 2001 Sun Microsystems, Inc.
* Portions Copyright 1999-2001 Language Technologies Institute,
* Carnegie Mellon University.
* All Rights Reserved. Use is subject to license terms.
*
* Permission is hereby granted, free of charge, to use and distribute
* this software and its documentation without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of this work, and to
* permit persons to whom this work is furnished to do so, subject to
* the following conditions:
*
* 1. The code must retain the above copyright notice, this list of
* conditions and the following disclaimer.
* 2. Any modifications must be clearly marked as such.
* 3. Original authors' names are not deleted.
* 4. The authors' names are not used to endorse or promote products
* derived from this software without specific prior written
* permission.
*
* DFKI GMBH AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH
* REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL DFKI GMBH NOR THE
* CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL
* DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
* PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF
* THIS SOFTWARE.
*/
package marytts.tools.voiceimport;
import java.io.ByteArrayOutputStream;
import java.io.File;
import java.io.IOException;
import java.text.DecimalFormat;
import java.util.Properties;
import java.util.SortedMap;
import java.util.TreeMap;
import marytts.unitselection.data.MCepDatagram;
import marytts.util.data.ESTTrackReader;
/**
* The mcepTimelineMaker class takes a database root directory and a list of basenames, and converts the related wav files into a
* mcep timeline in Mary format.
*
* @author sacha
*/
public class MCepTimelineMaker extends VoiceImportComponent {
protected DatabaseLayout db = null;
protected int percent = 0;
protected String mcepExt = ".mcep";
public final String MCEPDIR = "MCepTimelineMaker.mcepDir";
public final String MCEPTIMELINE = "MCepTimelineMaker.mcepTimeline";
public String getName() {
return "MCepTimelineMaker";
}
public SortedMap<String, String> getDefaultProps(DatabaseLayout theDb) {
this.db = theDb;
if (props == null) {
props = new TreeMap<String, String>();
props.put(MCEPDIR, db.getProp(db.ROOTDIR) + "mcep" + System.getProperty("file.separator"));
props.put(MCEPTIMELINE, db.getProp(db.FILEDIR) + "timeline_mcep" + db.getProp(db.MARYEXT));
}
return props;
}
protected void setupHelp() {
props2Help = new TreeMap<String, String>();
props2Help.put(MCEPDIR, "directory containing the mcep files");
props2Help.put(MCEPTIMELINE, "file containing all mcep files. Will be created by this module");
}
/**
* Reads and concatenates a list of mcep EST tracks into one single timeline file.
*
*/
public boolean compute() {
System.out.println("---- Importing Mel Cepstrum coefficients\n\n");
System.out.println("Base directory: " + db.getProp(db.ROOTDIR) + "\n");
/* Export the basename list into an array of strings */
String[] baseNameArray = bnl.getListAsArray();
/* Prepare the output directory for the timelines if it does not exist */
File timelineDir = new File(db.getProp(db.FILEDIR));
try {
/*
* 1) Determine the reference sampling rate as being the sample rate of the first encountered wav file
*/
WavReader wav = new WavReader(db.getProp(db.WAVDIR) + baseNameArray[0] + db.getProp(db.WAVEXT));
int globSampleRate = wav.getSampleRate();
System.out.println("---- Detected a global sample rate of: [" + globSampleRate + "] Hz.");
/* 2) Scan all the EST Mel Cepstrum Track files for min, max and total timeline duration */
System.out.println("---- Scanning for MCep min and MCep max...");
ESTTrackReader mcepFile; // Structure that holds the mcep track data
float[] current; // local [min,max] vector for the current mcep track file
float mcepMin, mcepMax, mcepRange; // Global min/max/range values for the mcep coefficients
float totalDuration = 0.0f; // Accumulator for the total timeline duration
long numDatagrams = 0l; // Total number of mcep datagrams in the timeline file
int numMCep = 0; // Number of mcep channels, assumed from the first mcep file
/* Initialize with the first file: */
/* - open and load */
// System.out.println( baseNameArray[0] );
File first = new File(getProp(MCEPDIR) + baseNameArray[0] + mcepExt);
if (!first.canRead()) {
throw new IOException("File " + first.getAbsolutePath() + " not readable!");
}
mcepFile = new ESTTrackReader(first.getAbsolutePath());
/* - get the min and the max */
current = mcepFile.getMinMax();
mcepMin = current[0];
mcepMax = current[1];
/* - accumulate the file duration */
totalDuration += mcepFile.getTimeSpan();
/* - accumulate the number of datagrams: */
numDatagrams += mcepFile.getNumFrames();
/* - get the number of mcep channels: */
numMCep = mcepFile.getNumChannels();
System.out.println("Assuming that the number of Mel Cepstrum coefficients is: [" + numMCep + "] coefficients.");
/* Then, browse the remaining files: */
for (int i = 1; i < baseNameArray.length; i++) {
percent = 50 * i / baseNameArray.length;
/* - open+load */
// System.out.println( baseNameArray[i] );
mcepFile = new ESTTrackReader(getProp(MCEPDIR) + baseNameArray[i] + mcepExt);
/* - get min and max */
current = mcepFile.getMinMax();
if (current[0] < mcepMin) {
mcepMin = current[0];
}
if (current[1] > mcepMax) {
mcepMax = current[1];
}
/* - accumulate and approximate of the total speech duration (to build the index) */
totalDuration += mcepFile.getTimeSpan();
/* - accumulate the number of datagrams: */
numDatagrams += mcepFile.getNumFrames();
}
mcepRange = mcepMax - mcepMin;
/*
* NOTE: accumulating the total mcep timeline duration (which is necessary for dimensioning the index) from the mcep
* track times is slightly more imprecise than accumulating durations from the residuals, but it avoids another loop
* through on-disk files.
*/
System.out.println("mcepMin = " + mcepMin);
System.out.println("mcepMax = " + mcepMax);
System.out.println("mcepRange = " + mcepRange);
System.out.println("---- Done.");
System.out.println("---- Translating the EST MCep tracks...");
/* 3) Open the destination timeline file */
/* Make the file name */
System.out.println("Will create the mcep timeline in file [" + getProp(MCEPTIMELINE) + "].");
/* An example of processing header: */
Properties props = new Properties();
String hdrCmdLine = "\n$ESTDIR/bin/sig2fv "
+ "-window_type hamming -factor 2.5 -otype est_binary -coefs melcep -melcep_order 12 -fbank_order 24 -shift 0.01 -preemph 0.97 "
+ "-pm PITCHMARKFILE.pm -o melcepDir/mcepFile.mcep WAVDIR/WAVFILE.wav\n";
props.setProperty("command", hdrCmdLine);
props.setProperty("mcep.order", String.valueOf(numMCep));
props.setProperty("mcep.min", String.valueOf(mcepMin));
props.setProperty("mcep.range", String.valueOf(mcepRange));
ByteArrayOutputStream baos = new ByteArrayOutputStream();
props.store(baos, null);
String processingHeader = baos.toString("latin1");
/* Instantiate the TimelineWriter: */
TimelineWriter mcepTimeline = new TimelineWriter(getProp(MCEPTIMELINE), processingHeader, globSampleRate, 0.1);
/* 4) Write the datagrams and feed the index */
long totalTime = 0l;
/* For each EST track file: */
for (int i = 0; i < baseNameArray.length; i++) {
percent = 50 + 50 * i / baseNameArray.length;
/* - open+load */
System.out.println(baseNameArray[i]);
mcepFile = new ESTTrackReader(getProp(MCEPDIR) + baseNameArray[i] + mcepExt);
wav = new WavReader(db.getProp(db.WAVDIR) + baseNameArray[i] + db.getProp(db.WAVEXT));
/* - For each frame in the mcep file: */
int frameStart = 0;
int frameEnd = 0;
int duration = 0;
long localTime = 0l;
for (int f = 0; f < mcepFile.getNumFrames(); f++) {
/* Get the datagram duration */
frameStart = frameEnd;
frameEnd = (int) ((double) mcepFile.getTime(f) * (double) (globSampleRate));
duration = frameEnd - frameStart;
/* NOTE: quantization is no more performed below, code&comments kept for archiving. */
/* Quantize the mcep coeffs: */
// short[] quantizedFrame = General.quantize( mcepFile.getFrame( f ), mcepMin, mcepRange );
/* Make a datagram from the quantized mcep coefficients: */
/*
* for ( int k = 0; k < quantizedFrame.length; k++ ) { datagramContents.writeShort( quantizedFrame[k] ); }
*/
/* Feed the datagram to the timeline */
mcepTimeline.feed(new MCepDatagram(duration, mcepFile.getFrame(f)), globSampleRate);
totalTime += duration;
localTime += duration;
}
// System.out.println( baseNameArray[i] + " -> mcep file says [" + localTime + "] samples, wav file says ["+
// wav.getNumSamples() + "] samples." );
}
mcepTimeline.close();
System.out.println("---- Done.");
/* 7) Print some stats and close the file */
System.out.println("---- mcep timeline result:");
System.out.println("Number of files scanned: " + baseNameArray.length);
System.out.println("Total speech duration: [" + totalTime + "] samples / ["
+ ((double) (totalTime) / (double) (globSampleRate)) + "] seconds.");
System.out.println("(Speech duration approximated from EST Track float times: [" + totalDuration + "] seconds.)");
System.out.println("Number of frames: [" + numDatagrams + "].");
System.out.println("Size of the index: [" + mcepTimeline.getIndex().getNumIdx() + "] ("
+ (mcepTimeline.getIndex().getNumIdx() * 16) + " bytes, i.e. "
+ new DecimalFormat("#.##").format((double) (mcepTimeline.getIndex().getNumIdx()) * 16.0 / 1048576.0)
+ " megs).");
System.out.println("---- mcep timeline done.");
} catch (SecurityException e) {
System.err.println("Error: you don't have write access to the target database directory.");
return false;
} catch (Exception e) {
e.printStackTrace();
System.err.println(e);
return false;
}
return (true);
}
/**
* Provide the progress of computation, in percent, or -1 if that feature is not implemented.
*
* @return -1 if not implemented, or an integer between 0 and 100.
*/
public int getProgress() {
return percent;
}
}