alphacep/unimrcp-vosk-plugin

Open source cross-platform implementation of MRCP protocol

unimrcp-vosk-plugin's Issues

Add GPU batch decoder support

Failed to Load DSO when ./unimrcpserver is started

Hi,

I am a total beginner and trying to establish UniMRCP server with the Vosk plugin.
I started the server in my local but having a linking issue below:

2022-07-27 10:16:08:539422 [INFO] Load Plugin [Vosk-Recog-1] [/usr/local/unimrcp/plugin/voskrecog.so]
2022-07-27 10:16:08:548585 [WARN] Failed to Load DSO: /usr/local/unimrcp/plugin/voskrecog.so: undefined symbol: vosk_recognizer_set_nlsml

I started the ./unimrcpclient, it is integrating with the server but I can not use the recognition engine. Rest works fine with demo plugins but when I try run recog it gets the error below:

2022-07-27 10:27:01:680069 [INFO] Receive SIP Event [nua_r_invite] Status 406 Not Acceptable [SIP-Agent-1]
2022-07-27 10:27:01:680110 [INFO] Receive SIP Event [nua_i_state] Status 406 Not Acceptable [SIP-Agent-1]

I tried to run ./bootstrap, ./configure, make, make install on unimrcp_vosk_plugin folder again and again nothing changed. My vosk_api is also installed on /home/username/vosk-api and I configured the file Makefile.am in plugins/vosk-recog as VOSK_HOME = /home/username/vosk-api

Could you explain what is wrong? Thanks.

Unimrcp with Vosk's Turkish model

When I try to use Vosk's Turkish model with unimrcp, I get the following result.

When I try any audio file, I always get the output of noinput.

I also tried the audio files that I used in the Vosk's Turkish model, in the Vosk's English model. It gave me a result, albeit meaningless.
Can you help me to use the Vosk's Turkish model in unimrcp?

Configure recognizer parameters in config file

Configure model path, nbest, etc.

Error Detected Noinput on the remote server

Hi!
I'm trying to use unimrcp-vosk-plugin and having error on the server side:
2022-02-06 00:17:23:382691 [INFO] Detected Noinput <1f025cda86e211ec@speechrecog>

Server-machine configurations are good in case of open ports and etc, as I have a connection messages between the server and the client. I'm working with:

• the default repo code;
• examples from .../unimrcp/data folder;
• default image from the docker-hub;

The issue happens while remote usage only. If the client and the server beside the same local network everything works.

I've already tried to set additional time-out parameters, because thought may be this is the long response issue, but it seems like it's not the case. Ive used:

• No-Input-Timeout: 15000
• Recognition-Timeout: 20000
• Speech-Complete-Timeout: 9000
• DTMF-Term-Timeout: 3000
• DTMF-Interdigit-Timeout: 9500

My docker port settings are:

• udp,tcp: 8060,5060
• udp: 8062, 4000-5000, 5000-6000
• tcp: 1544,1554,8000,4900

I've used the default configuration parameters but interchanged with my server public IP only.
And docker launch command is:
docker run -p 5000-6000:5000-6000/udp -p 1544:1544/tcp -p 1554:1554/tcp -p 8060:8060/udp -p 8060:8060/tcp -p 8062:8062/udp -p 4000-5000:4000-5000/udp -p 8000:8000/tcp -p 5060:5060/udp -p 4900:4900/tcp -p 554:554/tcp alphacep/vosk-unimrcp

Would you mind to tell, are there any options to overcome this? And how to do it in details?
Thx, in advance

if it arrive no input timeout ,it will send null result

Vosk API returns 'segmentation fault' from time to time

Gdb debugging results:

Memory leak:

I encountered a new error in a session:

pure virtual method called
terminate called without an active exception
Aborted (core dumped)

Issue with dtmf recognition in using unimrcp vosk plugin

Hi
I have been working on freeswitch / unimrcp / kaldi stack using vosk server plugin.
The speech recognition is working flawalessly however I am running into issue with dtmf recognition. I always get 002 no input timeout as my termination cause. I can replicate the issue using freeswitch or using the umc client and they both behave exactly same. In the unimrcp server logs, it shows the start and end events are being recognized for the each digit however the utterance files are empty. It appears as if the data is not being streamed to the mrcp server.

Here are the log files from unimrcpserver log

MRCP/2.0 203 RECOGNIZE 1
Channel-Identifier: 465b528207ee48df@speechrecog
Content-Id: request1@form-level
Content-Type: text/uri-list
Cancel-If-Queue: false
Content-Length: 19

builtin:dtmf/digits
2022-02-18 03:26:44:350231 [INFO] Assign Control Channel 465b528207ee48df@speechrecog to Connection 127.0.1.1:1544 <-> 127.0.0.1:33236 [0] -> [1]
2022-02-18 03:26:44:350563 [INFO] Process RECOGNIZE Request 465b528207ee48df@speechrecog [1]
2022-02-18 03:26:44:350856 [INFO] Open Utterance Output File [/usr/local/unimrcp/var/utter-8kHz-465b528207ee48df.pcm] for Writing
LOG (VoskAPI:CompileLooped():nnet-compile-looped.cc:345) Spent 0.00898099 seconds in looped compilation.
2022-02-18 03:26:44:409612 [INFO] Process RECOGNIZE Response 465b528207ee48df@speechrecog [1]
2022-02-18 03:26:44:409648 [INFO] State Transition IDLE -> RECOGNIZING 465b528207ee48df@speechrecog
2022-02-18 03:26:44:409722 [INFO] Send MRCPv2 Data 127.0.1.1:1544 <-> 127.0.0.1:33236 [83 bytes]
MRCP/2.0 83 1 200 IN-PROGRESS
Channel-Identifier: 465b528207ee48df@speechrecog

2022-02-18 03:26:44:498216 [INFO] Detected Start of Event 465b528207ee48df@speechrecog id:1
2022-02-18 03:26:44:541509 [INFO] Detected End of Event 465b528207ee48df@speechrecog id:1 duration:560 ts
2022-02-18 03:26:44:602958 [INFO] Detected Start of Event 465b528207ee48df@speechrecog id:2
2022-02-18 03:26:44:660659 [INFO] Detected End of Event 465b528207ee48df@speechrecog id:2 duration:560 ts
2022-02-18 03:26:44:724671 [INFO] Detected Start of Event 465b528207ee48df@speechrecog id:3
2022-02-18 03:26:44:782711 [INFO] Detected End of Event 465b528207ee48df@speechrecog id:3 duration:560 ts
2022-02-18 03:26:44:843399 [INFO] Detected Start of Event 465b528207ee48df@speechrecog id:4
2022-02-18 03:26:44:901544 [INFO] Detected End of Event 465b528207ee48df@speechrecog id:4 duration:560 ts
2022-02-18 03:26:49:444932 [INFO] Detected Noinput 465b528207ee48df@speechrecog
2022-02-18 03:26:49:445073 [INFO] Process RECOGNITION-COMPLETE Event 465b528207ee48df@speechrecog [1]
2022-02-18 03:26:49:445091 [INFO] State Transition RECOGNIZING -> RECOGNIZED 465b528207ee48df@speechrecog
2022-02-18 03:26:49:445184 [INFO] Send MRCPv2 Data 127.0.1.1:1544 <-> 127.0.0.1:33236 [138 bytes]
MRCP/2.0 138 RECOGNITION-COMPLETE 1 COMPLETE
Channel-Identifier: 465b528207ee48df@speechrecog
Completion-Cause: 002 no-input-timeout

I have tried matching the codecs between both client and server but that didn't make any difference.
let me know if you need any other debug information.

Vosk API returns 'segmentation fault'

Gdb debugging results:



Can you help me solve this problem?

During the test play_and_detect_speech, the playback audio is not complete, it will be interrupted

During the test play_and_detect_speech, the playback audio is not complete, it will be interrupted

plugin-factory + engine config after installation?

Hi, this is the resulting configuration after my installation; kindly let me know of the following;

• what should we put in the plugin-factory section to enable the plugin ?
• does port 2700 actually tells UniMRCP where to connect to use the plugin?
• are there any other parameters available we can use in the params section?
• is there any doc somewhere we can look for to correctly configure the plugin?
  Thanks!

   <plugin-factory>
     <engine id="Demo-Synth-1" name="demosynth" enable="true"/>
     <engine id="Demo-Recog-1" name="demorecog" enable="true"/>
     <engine id="Demo-Verifier-1" name="demoverifier" enable="true"/>
     <engine id="Recorder-1" name="mrcprecorder" enable="true"/>

       <engine>
               <id>Vosk</id>
               <name>Vosk Engine</name>
               <description>Vosk Speech Recognition Engine</description>
               <!-- <module_path>/path/to/vosk-engine</module_path> -->
               <params>
                       <param name="server" value="192.168.0.57"/>
                       <param name="port" value="2700"/>
                       <!-- Add any additional parameters here -->
               </params>
       </engine>
   </plugin-factory>
 </components>

Return NSXML

Clients have trouble to parse json

Freeswitch grammar

Hi Everyone,
I have a question about grammar implemetation. If i send a grammar from freeswitch in the MRCP DEFINE-GRAMMAR method, does it has any effect on the result ?
Does vosk plugin takes the grammar sent from the freeswitch into account ?

Any help is appreciated, thanks.

How to setting chinese asr model

I have compiled unimrcp-vosk-plugin,
and how to set Chinese ASR model?
Thanks!