package edu.cmu.sphinx.demo.speakerid;
import java.io.IOException;
import java.net.URL;
import java.util.ArrayList;
import edu.cmu.sphinx.api.Configuration;
import edu.cmu.sphinx.api.SpeechResult;
import edu.cmu.sphinx.api.StreamSpeechRecognizer;
import edu.cmu.sphinx.decoder.adaptation.Stats;
import edu.cmu.sphinx.decoder.adaptation.Transform;
import edu.cmu.sphinx.speakerid.Segment;
import edu.cmu.sphinx.speakerid.SpeakerCluster;
import edu.cmu.sphinx.speakerid.SpeakerIdentification;
import edu.cmu.sphinx.util.TimeFrame;
public class SpeakerIdentificationDemo {
/**
* Returns string version of the given time in miliseconds
*
* @param seconds
* @return time in format mm:ss
*/
public static String time(int seconds) {
return (seconds / 60000) + ":"
+ (Math.round((double) (seconds % 60000) / 1000));
}
/**
*
* @param speakers
* An array of clusters for which it is needed to be printed the
* speakers intervals
* @throws IOException
*/
public static void printSpeakerIntervals(
ArrayList<SpeakerCluster> speakers, String fileName)
throws IOException {
int idx = 0;
for (SpeakerCluster spk : speakers) {
idx++;
ArrayList<Segment> segments = spk.getSpeakerIntervals();
for (Segment seg : segments)
System.out.println(fileName + " " + " "
+ time(seg.getStartTime()) + " "
+ time(seg.getLength()) + " Speaker" + idx);
}
}
/**
* @param speakers
* An array of clusters for which it is needed to get the
* speakers intervals for decoding with per-speaker adaptation
* with diarization.
* @param url
* Url for the audio
* @throws Exception
*/
public static void speakerAdaptiveDecoding(ArrayList<SpeakerCluster> speakers,
URL url) throws Exception {
Configuration configuration = new Configuration();
// Load model from the jar
configuration
.setAcousticModelPath("resource:/edu/cmu/sphinx/models/acoustic/wsj");
configuration
.setDictionaryPath("resource:/edu/cmu/sphinx/models/acoustic/wsj/dict/cmudict.0.6d");
configuration
.setLanguageModelPath("resource:/edu/cmu/sphinx/models/language/en-us.lm.dmp");
StreamSpeechRecognizer recognizer = new StreamSpeechRecognizer(
configuration);
TimeFrame t;
SpeechResult result;
for (SpeakerCluster spk : speakers) {
Stats stats = recognizer.createStats(1);
ArrayList<Segment> segments = spk.getSpeakerIntervals();
for (Segment s : segments) {
long startTime = s.getStartTime();
long endTime = s.getStartTime() + s.getLength();
t = new TimeFrame(startTime, endTime);
recognizer.startRecognition(url.openStream(), t);
while ((result = recognizer.getResult()) != null) {
stats.collect(result);
}
recognizer.stopRecognition();
}
Transform profile;
// Create the Transformation
profile = stats.createTransform();
recognizer.setTransform(profile);
for (Segment seg : segments) {
long startTime = seg.getStartTime();
long endTime = seg.getStartTime() + seg.getLength();
t = new TimeFrame(startTime, endTime);
// Decode again with updated SpeakerProfile
recognizer.startRecognition(url.openStream(), t);
while ((result = recognizer.getResult()) != null) {
System.out.format("Hypothesis: %s\n",
result.getHypothesis());
}
recognizer.stopRecognition();
}
}
}
public static void main(String[] args) throws Exception {
SpeakerIdentification sd = new SpeakerIdentification();
URL url = SpeakerIdentificationDemo.class.getResource("test.wav");
ArrayList<SpeakerCluster> clusters = sd.cluster(url.openStream());
printSpeakerIntervals(clusters, url.getPath());
speakerAdaptiveDecoding(clusters, url);
}
}