Sphinx4 within MaxMSP

[pasquetje]

Hello,

I have started to implement Sphinx4 in Max/MSP ( http://www.cycling74.com ).

---

• Max takes the sound input as in[i].

I'm trying to get rid of the Microphone in Sphinx4 and use in[i] (with AudioData()?).

QUESTION : How can I get rid of the stream from the system (microphone) and replace it by an audio stream (the one from Max) in the form of in[i]?

Thank you!

[pasquetje]

hi!

I'm still trying to have something like ( ins[0].vec is the vector from the live sound input ):

method {
int i;
float[] in = ins[0].vec;

for(i = 0; i < in.length;i++)
{ make the Sphinx4 audio stream input in here with in[i] }
}

thank you very much!

[kiuzzi]

Hi

Have a look at the Transcriber Demo Program shipped with Sphinx4.

Good Luck

[pasquetje]

In the Transcriber demo, you have

AudioInputStream ais = AudioSystem.getAudioInputStream(audioURL);

With audioURL is a sound file.
The max input is a stream of float values as a stream of data (realtime). I need something like:

method {
int i;
float[] in = ins[0].vec;

for(i = 0; i < in.length;i++)
{ make the Sphinx4 audio stream input in here with in[i] }
}

microphone.java does the job but it uses AudioInputStream and AudioSystem.
I looked in StreamDAtaSource but...

short[i] shortin = (short)in[i];

StreamDataSource reader = (StreamDataSource) cm.lookup("streamDataSource");
reader.setInputStream( new AudioInputStream(shortin), "streamName");

[pasquetje]

Actually, "make the Sphinx4 audio stream input in here with in[i]" is maybe wrong.

I just need to feed the AudioInputStream from the samples coming from in[] (where 'in[]' has floats).
The loop can be used to get input samples one by one inside a single dsp vector size.