And I got a perfect result, however when I run the file included iwth the WavFile demo, 12345.wav through this config I get "what you eat or five", which is not so good. I am partially worried about the audio formats I am using, so I used some software to convert the an4 data to a wavfile and when I run that through my java code (see the routines below), I get the following:
File is in :PCM_SIGNED 44100.0 Hz, 16 bit, stereo, 4 bytes/frame, little-endian
for the audio format. This file STILL PRODUCES PERFECT RESULTS despite the fact that the format is all wrong. however when I attempt to convert it to the right format using (see java code below):
which should down convert it to match my config.xml file, I get no results whatsoever.
So any ideas as to what is going on here, why am I able to work with a 44.1 khz file and not with a 16khz file, when I specified a 16khz file in the config.xml file? Do all sphinx inputs need to be mono, and how do I set the bytesPerRead property (how is this calculated)?
I see now that java can't down covert my file to 16khz, but I still don't understand how the file actually works with the setup in the config.xml file.
My goal is to pass a converted mp3 file to sphinx and have it spit out some meaningful text, currently I am testing it with wavs and getting 0 words correct in a 10 second file using hub4. I must be doing something wrong?
Please Help!!!!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
All-
As my previous post mentioned I have been having problems converting the transcribe config files to work with text data.
I decided to run some tests using the an4 dataset, specifically the:
an4_clstk/fbbh/cen8-fbbh-b.raw
audio file.
I ran this through the following congif.xml:
<?xml version="1.0" encoding="UTF-8"?>
<!--
Sphinx-4 Configuration file
-->
<!-- ******** -->
<!-- biship configuration file -->
<!-- ******** -->
<config>
<!-- ******** -->
<!-- frequently tuned properties -->
<!-- ******** -->
<property name="absoluteBeamWidth" value="1000"/>
<property name="relativeBeamWidth" value="1E-80"/>
<property name="absoluteWordBeamWidth" value="500"/>
<property name="relativeWordBeamWidth" value="1E-60"/>
<property name="wordInsertionProbability" value="1E-16"/>
<property name="languageWeight" value="7.0"/>
<property name="silenceInsertionProbability" value=".1"/>
<property name="frontend" value="mfcFrontEnd"/>
<property name="recognizer" value="recognizer"/>
<property name="showCreations" value="false"/>
</config>
And I got a perfect result, however when I run the file included iwth the WavFile demo, 12345.wav through this config I get "what you eat or five", which is not so good. I am partially worried about the audio formats I am using, so I used some software to convert the an4 data to a wavfile and when I run that through my java code (see the routines below), I get the following:
File is in :PCM_SIGNED 44100.0 Hz, 16 bit, stereo, 4 bytes/frame, little-endian
for the audio format. This file STILL PRODUCES PERFECT RESULTS despite the fact that the format is all wrong. however when I attempt to convert it to the right format using (see java code below):
AudioFormat.Encoding.PCM_SIGNED,
16000,
16,
1,
2,
2,
16000,
false );
which should down convert it to match my config.xml file, I get no results whatsoever.
So any ideas as to what is going on here, why am I able to work with a 44.1 khz file and not with a 16khz file, when I specified a 16khz file in the config.xml file? Do all sphinx inputs need to be mono, and how do I set the bytesPerRead property (how is this calculated)?
Very confused,
James
Here is all of my java code:
private AudioInputStream convertFormat(AudioInputStream audioInputStream) {
AudioFormat audioFormat = audioInputStream.getFormat();
System.out.println( "Play input audio format=" + audioFormat );
// Convert compressed audio data to uncompressed PCM format. // if ( audioFormat.getEncoding() != AudioFormat.Encoding.PCM_SIGNED ) { // if ((audioFormat.getEncoding() != AudioFormat.Encoding.PCM) || // (audioFormat.getEncoding() == AudioFormat.Encoding.ALAW) || // (audioFormat.getEncoding() == AudioFormat.Encoding.MP3)) { AudioFormat newFormat = new AudioFormat( AudioFormat.Encoding.PCM_SIGNED, 16000, 16, 1, //audioFormat.getChannels(), //audioFormat.getChannels() * 2, 2, 16000, false ); System.out.println( "Converting audio format to " + newFormat ); AudioInputStream newStream = AudioSystem.getAudioInputStream( newFormat, audioInputStream ); audioFormat = newFormat; audioInputStream = newStream; //} return audioInputStream; } public void parse(URL audioFileURL) { try { //URL configURL = PodCastParser.class.getResource("hub4.config.xml"); //URL configURL = PodCastParser.class.getResource("myWSJconfig.xml"); //URL configURL = PodCastParser.class.getResource("myhub4config.xml"); URL configURL = PodCastParser.class.getResource("transcriberconfig.xml"); // URL configURL = PodCastParser.class.getResource("hellongram.config.xml"); System.out.println("Loading Recognizer...\n"); ConfigurationManager cm = new ConfigurationManager(configURL); //Recognizer recognizer = (Recognizer) cm.lookup("wordRecognizer"); Recognizer recognizer = (Recognizer) cm.lookup("recognizer"); /* allocate the resource necessary for the recognizer */ recognizer.allocate(); System.out.println("Decoding " + audioFileURL.getFile()); //System.out.println(AudioSystem.getAudioFileFormat(audioFileURL)); StreamDataSource reader = (StreamDataSource) cm.lookup("streamDataSource"); InputStream is; try { is = AudioSystem.getAudioInputStream(audioFileURL); System.out.println("File is in :"+((AudioInputStream) is).getFormat()); is = convertFormat((AudioInputStream)is); } catch (UnsupportedAudioFileException uafe) { System.out.println ("Reading " + audioFileURL.toExternalForm() + " as raw audio file."); is = new FileInputStream(audioFileURL.getFile()); } /* set the stream data source to read from the audio file */ reader.setInputStream(is, audioFileURL.getFile()); /* decode the audio file */ boolean done = false; while (!done) { /* * This while loop will terminate after the last utterance * in the audio file has been decoded, in which case the * recognizer will return null. */ Result result = recognizer.recognize(); if (result != null) { String resultText = result.getBestResultNoFiller(); System.out.println(result.getTimedBestResult(true, true)); System.out.println(resultText); } else { done = true; } }// Result result = recognizer.recognize();
// String resultText = result.getBestResultNoFiller();
// System.out.println("You said: " + resultText + "\n");
//
I see now that java can't down covert my file to 16khz, but I still don't understand how the file actually works with the setup in the config.xml file.
My goal is to pass a converted mp3 file to sphinx and have it spit out some meaningful text, currently I am testing it with wavs and getting 0 words correct in a 10 second file using hub4. I must be doing something wrong?
Please Help!!!!