Run Julian Live

1. First you need to create your Julian configuration file. Copy this sample configuration file (julian.jconf) to you 'voxforge/auto' folder. For details on the parameters contained in the julian.jconf file, see the Julius manual. The main parameters are shown below:

-dfa sample.dfa
-v sample.dict
-h acoustic_model_files/hmmdefs
-hlist acoustic_model_files/tiedlist
-smpFreq 48000

2. Make sure your Microphone Volume is at a setting that is similar to when you created your audio files.

3. Then run Julian with:

$cd voxforge/auto
$julian -input mic -C julian.jconf

4. The Julian will display its startup information on your screen. You can begin speaking when you see:

------------- System Info end -------------

        ************************************************************
        * NOTICE: The first input may not be correctly recoginized *
        *         since no CMN parameter is available on startup. *
        ************************************************************

------
### read waveform input
<<< please speak >>>

The first 2-3 seconds of your speech will not be recognized as Julian adjusts its recognition levels. You may need to adjust your volume to get better recognition results.

You should get fair recognition results. To improve recognition, your Acoustic Model needs more audio training data. You need to create new prompts, and record more speech audio files based on these prompts in order to create better Acoustic Models. However, with an automated script, it is much easier to record additional audio data.

To add more speech audio to your Acoustic Model, keep the audio and prompts you have already created, and follow the steps in this how-to:

add your new prompts to the prompts file you created in Step 2,
record new audio files corresponding to the your prompts as shown in Step 3, and
Update your codetrain.scp file to include your new audio files, and re-run the training script as shown in Step 4.

As you add more audio data, you will begin to notice improvements in your recognition accuracy, and it will reach a point where you no longer need to add any additional training data.

The steps in this how-to are for the creation of a single user Acoustic Model. To create a multi-user Acoustic Models, simply get other people to contribute audio as shown in the previous steps, and re-run your script.

Comments

Click the 'Add' link to add a comment to this page; click the 'Read More' link to view replies to a posted comment.

Add • Search

Step 7 Fixing the silence model

By ale.78 - 6/18/2010 - 1 Replies

Hi, I'm following the tutorial and using Windows vista, Cygwin, HTK3.3, julian by julius 3.5.2.

When I run HTK_Compile_Model.sh it works well till step 6. Then on step 7 - Fixing the silence model - it stops to work and remains stopped showing making hmm4.

Any hint? any step made wrong? something to check?

Do you need some more information?

Please help...

It doesn't work on my Ubuntu Karmic

By Leonardo - 11/24/2009 - 1 Replies

When I execute ./julian -input mic -C julian.jconf it results in this:

include config: julian.jconf
###### check configurations
###### initialize input device
adin_mic_standby: device=/dev/dsp
adin_mic_standby: open device: Device or resource busy
Error: failed to ready input device
Terminated

If I run lspci -vvv, the audio device info is this:

00:1b.0 Audio device: Intel Corporation 82801I (ICH9 Family) HD Audio Controller (rev 03)
    Subsystem: Toshiba America Info Systems Device ff1e
    Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
    Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
    Latency: 0, Cache Line Size: 64 bytes
    Interrupt: pin A routed to IRQ 22
    Region 0: Memory at d6800000 (64-bit, non-prefetchable) [size=16K]
    Capabilities: <access denied>
    Kernel driver in use: HDA Intel
    Kernel modules: snd-hda-intel

Any workaround?.

Leonardo.


Username	Password