Comments

Hi,

I am getting a error at time of decoding using sphinx3_decode.

INFO: utt.c(195): Processing: 1
INFO: feat.c(1148): At directory /home/biswajit/Desktop/Tutorial/Decode/feat
INFO: feat.c(378): Reading mfc file: '/home/biswajit/Desktop/Tutorial/Decode/feat/1.mfc'[0..-1]
INFO: cmn.c(175): CMN:  9.10 -0.64 -0.16 -0.09 -0.14 -0.01 -0.17 -0.09 -0.10 -0.15 -0.13 -0.08 -0.17
....................................
WARNING: "vithist.c", line 784: No word exit in frame 356, using exits from frame 2
ERROR: "vithist.c", line 811: No word exit in frame 356, using exits from frame 2
INFO: fast_algo_struct.c(397): HMMHist[0..0](1): 357(100)
INFO: lm.c(950):         0 tg(),         0 tgcache,        0 bo;     0 fills,        0 in mem (0.0%)
INFO: lm.c(954):        2 bg(),        2 bo;     0 fills,        0 in mem (0.0%)

Backtrace(1)
FV:1>                 WORD  SFrm  EFrm AScr(UnNorm)    LMScore  AScr+LScr     AScale
fv:1>                <sil>     0     2      -349443     -82114    -431557     451453
fv:1>                <sil>     3   356     69734465     -82114   69652351   69734465
FV:1>                TOTAL                 69385022    -164228

FWDVIT: (1)
FWDXCT: 1 S 70633636 T 69369672 A 69385022 L -15350 0 -349443 -7675 <sil> 3 69734465 -7675 <sil> 357

INFO: stat.c(156):  357 frm;    18 cdsen/fr,  144 cisen/fr,   126 cdgau/fr,  1105 cigau/fr, Sen 0.44, CPU 0.45 Clk [Ovrhd 0.39 CPU 0.39 Clk];      5 hmm/fr,   0 wd/fr, Search: 0.00 CPU 0.00 Clk (1) 
INFO: corpus.c(661): 1:    1.6 sec CPU,    1.6 sec Clk;  TOT:      1.6 sec CPU,      1.6 sec Clk

INFO: stat.c(206): SUMMARY:  357 fr;  17 cdsen/fr, 144 cisen/fr, 125 cdgau/fr, 1105 cigau/fr, 0.44 xCPU 0.45 xClk [Ovhrd 0.39 xCPU  0 xClk];  4 hmm/fr, 0 wd/fr, 0.00 xCPU 0.00 xClk;  tot: 0.44 xCPU, 0.45 xClk
biswajit  8150 73.8  0.2   7552  4952 pts/1    S+   15:34   0:03 sphinx3_decode -mdef /home/biswajit/Desktop/Tutorial/sruti/3s_8g_17_02_11/model_architecture/sruti.1000.mdef -senmgau .cont. -mean /home/biswajit/Desktop/Tutorial/sruti/3s_8g_17_02_11/model_parameters/sruti.cd_cont_1000_8/means -var /home/biswajit/Desktop/Tutorial/sruti/3s_8g_17_02_11/model_parameters/sruti.cd_cont_1000_8/variances -mixw /home/biswajit/Desktop/Tutorial/sruti/3s_8g_17_02_11/model_parameters/sruti.cd_cont_1000_8/mixture_weights -tmat /home/biswajit/Desktop/Tutorial/sruti/3s_8g_17_02_11/model_parameters/sruti.cd_cont_1000_8/transition_matrices -lw 10 -feat 1s_c_d_dd -beam 1e-80 -wbeam 1e-40 -dict /home/biswajit/Desktop/Tutorial/sruti/etc/sruti.dic -fdict /home/biswajit/Desktop/Tutorial/sruti/etc/sruti.filler -lm /home/biswajit/Desktop/Tutorial/sruti/etc/sruti.ug.lm.DMP -wip 0.2 -ctl /home/biswajit/Desktop/Tutorial/Decode/decode123.fields -ctloffset 0 -ctlcount 627 -cepdir /home/biswajit/Desktop/Tutorial/Decode/feat -cepext .mfc -hyp /home/biswajit/Desktop/Tutorial/Decode/hyp.txt -logfn /home/biswajit/Desktop/Tutorial/Decode/log -agc none -varnorm no -cmn current
biswajit  8151  0.0  0.0   5100   984 pts/1    S+   15:34   0:00 sh -c ps aguxwww | grep sphinx3_decode
biswajit  8153  0.0  0.0   4360   708 pts/1    S+   15:34   0:00 grep sphinx3_decode

When i am using normal model, there is no problem but with VTLN enabled model it is giving error.

Please suggest me some solution.

Hi all,

6 years ago, I wrote an article titled "Do we have a true open source dictation machine?".  At that time, there was no such thing called voxforge.  It is a programmer/speech researcher's dream to build open source speech recognizer.  

But that was impossible.  As I attempted to argue in the article, unless a recognizer open both the ASR-engine and the models. We could hardly call that open source speech recognizer.

VoxForge changed this.  From my initial browse, Ken had brought awareness of the community that we need to collect *both data and source code*.

VoxForge's effort has been benefiting several projects and I find this is an exciting future. 

Should I rewrite my article now?  My friends in the speech community! May be my view was outdated.

I don't know the answer, but kudos to everyone in VoxForge and notably Ken who started this effort.  I sincerely hope one day open source speech recognition will be avaiable to everyone in the planet. 

Arthur Chan

or

a nick I love

"The Grand Janitor of Sphinx"

i need an open source speech rec model for my business. I can contribute 100,000+ dictations and associated transcriptions to the project (medical reports with all patient and physician information redacted). i also need a very bright programmer/technical person who is into speech recognition to work for money to build a speech model for my company. this person would be responsible for managing the dictations and the contributions to the VoxForge project. I don't know a lot about open source - we are big users of several commercial applications. however, the commercial applications are not as good as they could be - but are all closed up black boxes, so we cannot get at them to do better - we just spend LOTS of money to get what we get. We know we can do better - we have a lot of technical resources - but we need a strong engineer who knows HMM and how to build a real application with it. anybody interested. drop me an email. totally confidential.

Hi there,

 Made a blog post on your wonderful project at http://flossexperiences.wordpress.com/2010/04/30/voxforge-speech-recognition/ . Hope you can address some of the concerns I have outlined or atleast have some bugs opened. I would have opened the bugs myself but only if openid would have been implemented at the site. 

Good luck with the project :)

In reponse to this post on LWN.net: Simon - speech activated user interface for KDE (KDE.News):

KDE.News has a look at simon, which is a speech-activated interface for KDE. It looks like an interesting project, but, unfortunately, may suffer from some licensing snags: "HTK, the toolkit responsible for the HMM [Hidden Markov Model] evaluation is distributed under GPL-incompatible, restrictive license that prevents redistribution. In order to install simon, one must separately download HTK from their website which requires registration. The source is available, [...]

bedhar (Simon developer) replies with:

How does the GPL license for voice data affect the owner's obligations with respect to application source code that uses the voices? In general, GPL licensed code "infects" any code that links to it. If my code requires voxforge voices or voices that are derived from voxforge voices to run, do I have to open source my application code as well as voices?

What is GPL and does the word "melarkia" have anything to do with it?  I did a google search for "melarkia" and it sent me to VoxForge.  Anyone know the answer?  Thanks.

This article provides an support for VoxForge's use of the GPL: The different reasons for company code contributions.   In it, the author states:

[...] licensing issues are the main reason for publishing back, but separated by very few percentage points other reasons appear: the signaling advantage (being good players), the R&D sharing, and many others. In this sense, my view is that the GPL creates an initial context (by forcing the publication of source code) that creates a secondary effect - reuse and quality improvement - that appears after some time. In fact, our research shows that companies need quite some time to grasp the advantages of reuse and participation; the GPL enforces participation for enough time that companies discovers the added benefits, and start moving their motivations to economic reasons, as compared to legal enforcing or legal risks.

pls help me in eliminating this error

Hello,

Congratulations for your project :)

My question is simple, why not use also LGPL?

If we could put this module in a commercial software, using LGPL would be mandatory to commit any modifications to the source code (that would bring more coders and investment into this module).

This would be fair for the community, it would bring more people and more investment, and wouldn't be anti-enterprise. Would be the best of both worlds.

Not all enterprises are evil. There are good enterprises that bring innovation, create jobs, pay taxes and make a good contribution to the world. The problem is when those companies become too greedy to share any innovation. Thats why LGPL is so nice.

Think about it :)

Cya 

Be very careful in the way you license this data. If you release a great collection of GPL data, and someone else releases a great collection of data under, for example, a CC-by license, it will probably be illegal to combine the two corpuses and make a working speech recognition product.

I have been involved in 2 projects that went through immense pain because of this issue. The solution is as follows:

1. Set up a proper legal entity to hold the data (an organisation).
2. Make sure the organisation has good governance that end users will approve of.
3. Make end users assign a non-exculsive irrevokeable perpetual license to the organisation for the organisation do to whatever it democratically decides in future.

If you don't make this clear right from the start, any attempt to change the license, for whatever reason, will be impossible.

For example, if a court rules that using a subset of the data derived from VoxForge on a hand-held device doesn't comply with the source code rule, and you have to bundle gigabytes of data in order to use VoxForge derived engines on mobile devices, you will be back to square one, and will have to start collecting data all over again. Who knows why a court may do that, but unless you can be sure, you can't risk not assigning data.

The organisation could have in it's constitution, (terms of incorporation or whatever it's called in the relevant jurisdiction) that it will always make the data available under the GPL, but may additionally make it available under other licenses.

I think there is a misunderstanding of the difference between audio data and source code. When I create an executable from source code, I might modify the source code. But when I create acoustic models, I don't modify the data that is used to train the acoustic models. So, if I have to distribute the data along with my models, I'd be distributing an identical copy of the data. (Not to mention the difficulties of distributing gigabytes of data...)