ASR with uniMRCP

[bharani]

Hi arsen,



I have configured in my client freeswitch(192.168.1.15) and with
unimrcp server(192.168.1.18). Based on your instructions and somewhat
from this discussion group wat i hv learned about freeswitch and
uniMRCP stuff. Thank you very much..

I expecting you hv to clear and you ill be with me for forthcoming my
doubts, suggestions, idea, error and all any other.

Here, i have not use pizza demo application instead of i made a simple
application to recognize a single sentence only.... for tat i made
some changes in SpeechTools.js also i created a prabhu_asr.js file.

also i tested a simple ASR uniMRCP stuff in two different ways both
test i could not succeed.

The 1st way is in client side with mod_pocketsphinx module.

i pasted here the result from CLI both client FS and server uniMRCP.

the below CLI result from client FS

http://pastebin.com/m5a244525


the below CLI result from server uniMRCP

http://pastebin.com/m7be707f3

the recognition thing doesnt happen in this test.... I was wrong
somewhere.. please check out that verbose result and figure out where
i hv been wrong.

Then i thought why we to use pocketsphinx model in client side....
so, the 2nd way is i removed asr / tts mod_pocketsphinx from
modules.conf then i reinstalled freeswitch...

Just check this below link for this test result,

http://pastebin.com/m706a35b5

from this test case there is no result from uniMRCP CLI


moreover,Tell me brief about the following... What do u meant on it.

1. [ c:0 a:1 v:0 ]

2. ASR-0 , 1 , 2

BY

PRABHU P

[bharani]

from this test case there is no result from uniMRCP CLI.

Tell me where i made wrong.....

[bharani]

[Arsen Chaloyan]

Hi PRABHU,

On Sat, Sep 5, 2009 at 2:28 PM, bharani <prabh...@gmail.com> wrote:

Hi arsen,



I have configured in my client freeswitch(192.168.1.15) and with
unimrcp server(192.168.1.18). Based on your instructions and somewhat
from this discussion group wat i hv learned about freeswitch and
uniMRCP stuff. Thank you very much..

 

OK.

I expecting you hv to clear and you ill be with me for forthcoming my
doubts, suggestions, idea, error and all any other.

Yes, the discussion group is intended for similar discussions. Your thoughts and suggestions are always welcome.

Here, i have not use pizza demo application instead of i made a simple
application to recognize a single sentence only.... for tat i made
some changes in SpeechTools.js also i created a prabhu_asr.js file.

Well, you may want to try simple application first, however demo pizza is a funny one, my kids keep asking when we are going to order a pizza next time :)

also i tested a simple ASR uniMRCP stuff in two different ways both
test i could not succeed.

The 1st way is in client side with mod_pocketsphinx module.

i pasted here the result from CLI both client FS and server uniMRCP.

the below CLI result from client FS

http://pastebin.com/m5a244525


the below CLI result from server uniMRCP

http://pastebin.com/m7be707f3

Everything looks good, however try to explicitly specify your server IP address. Thus, if your server runs on 192.168.1.18, find "auto" in unimrcpserver.xml and replace with "192.168.1.18". I see from logs, that it defaults to localhost.

the recognition thing doesnt happen in this test....  I was wrong
somewhere.. please check out that verbose result and figure out where
i hv been wrong.

Then i thought why we to use pocketsphinx model in client side....
so, the 2nd way is  i removed asr / tts mod_pocketsphinx from
modules.conf then i reinstalled freeswitch...

You can keep other ASR/TTS modules in FreeSWITCH, you just need to specify which module to use from your application

Just check this below link for this test result,

http://pastebin.com/m706a35b5

from this test case there is no result from uniMRCP CLI.

Tell me where i made wrong.....

The only thing I've seen so far is wrong IP address, fix it and try again 

moreover,Tell me brief about the following... What do u meant on it.

1. [ c:0 a:1 v:0 ]

Number of control (MRCPv2) and audio and video media ("m=") is SDP

2. ASR-0 , 1 , 2

Not sure where this output is from.

BY

PRABHU P

--
Arsen Chaloyan
The author of UniMRCP
http://www.unimrcp.org

[bharani]

Hi Arsen,


i have done wat did u mentioned in ur previous reply....

Now, i am getting some different log results...

let me know, wat is the purposes of result.xml file locating in data
folder, also please tell me about "session:request1@form-
level.store". i am getting recognition result "one" when i spoke
"delivery or takeout".

and where exact location the recognition result will be store. tell me
the path, except CLI in the screen..

please pay attention below link

http://pastebin.com/m20cc3a03

and MRCP log is,

http://pastebin.com/m3fbc5876

and i could not save waveform of speech file configured in
pocketsphinx.xml

and i am getting a pcm file in data folder for each recognition test..
how do i listen tat file.. any conversion through sox?

awaiting for ur immediate reply

Thank you............

[Arsen Chaloyan]

Hi Prabhu,

I was offline during the day. If you post to the group there should be others available to help you out.
Anyway, I'm afraid you're using demo recognizer plugin, which just simulates the recognition regardless what you actually said. Demo recognizer uses result.xml to construct the recognition result.

To use pocketsphinx plugin you must enable it from unimrcpserver.xml

      <engine name="PocketSphinx-1" class="mrcppocketsphinx" enable="1"/>
      <engine name="Demo-Recog-1" class="demorecog" enable="0"/>

and pay attention to the output on start-up