Speech recognition demo : Failed to initialize ASR recognizer

[pmx]

Hi,

I am on ubuntu 18.04, working with Openvino 2020.1 and want to try the speech demos of this release. Following the doc, I have run the command <INSTALL_DIR>/deployment_tools/demo/demo_speech_recognition.sh to download models and build dependecies. At the end of this script, it automatically runs 2 demos, Inference Engine offline speech recognition demo, which runs succesfully, and Inference Engine real time speech recognition demo, which starts an GUI and on which I have the folllowing error messages when I click on one of the "Recognize" button :

- In the GUI:  Failed to initialize ASR recognizer,

- In the console: 

gx(0) [/opt/intel/openvino/data_processing/audio/speech_recognition/demos/live_speech_recognition_demo]
-> ./run_demo.sh 
[setupvars.sh] OpenVINO environment initialized
target_precision = FP32
Using model configuration file /home/gx/openvino_models/ir/intel/lspeech_s5_ext/FP32/speech_lib.cfg
Run Inference Engine live speech recognition demo

/home/gx/.local/lib/python3.6/site-packages/wx/lib/pubsub/__init__.py:25: wxPyDeprecationWarning: wx.lib.pubsub has been deprecated, plese migrate your code to use pypubsub, available on PyPI.
  wx.wxPyDeprecationWarning)

(speech_demo_app.py:24794): Gtk-WARNING **: 12:07:38.298: Failed to measure available space: Aucun chemin d'accès pour atteindre l'hôte cible
[ INFO ] Using feature transformation /home/gx/openvino_models/ir/intel/lspeech_s5_ext/FP32/lspeech_s5_ext.feature_transform
[ ERROR ] Could not parse number '-74.24161' in /home/gx/openvino_models/ir/intel/lspeech_s5_ext/FP32/lspeech_s5_ext.feature_transform
[ ERROR ] Add shift dimension mismatch (0 != 143)
2020-02-12 12:08:27,674 [ERROR] speech_engine.py::initialize | Failed to initialize engine (lib: 'libspeech_library.so'). Status code: -4
2020-02-12 12:08:27,675 [ERROR] speech_proxy.py::speech_process | Failed to initialize ASR recognizer
2020-02-12 12:08:27,674 [ERROR] speech_engine.py::initialize | Failed to initialize engine (lib: 'libspeech_library.so'). Status code: -4
2020-02-12 12:08:27,675 [ERROR] speech_proxy.py::speech_process | Failed to initialize ASR recognizer
2020-02-12 12:08:27,676 [ERROR] speech_manager.py::initialize | Failed to initialize Speech Library process
2020-02-12 12:08:27,706 [ERROR] main_controller.py::_recognize_wave_start | Failed to initialize ASR recognizer
 

Thanks for your help !

[pmx]

Still no answer for my post. Sorry to ask, I am not familiar with this forum. Does it mean that intel techs have no idea on my problem ? Or can it simply be that nobody has read my post yet ?

[Luis_at_Intel]

Hi pmx,

Thanks for reaching out. I have not encountered such issue before, I just tested the demo on Ubuntu 18.04 and everything seems to work fine. I have a couple of questions, did you run all the installation steps (installed OpenVINO dependencies sudo -E ./install_openvino_dependencies.sh, Set the Environment Variables, Configure the Model Optimizer, run the Verification Scripts)?

Please also provide more info about your environment (processor, Python* version, CMake* version, hardware plugin used i.e CPU, MYRIAD,etc)?

Regards,

Luis

[elios]

Hi,

I have the same problem as rms. I'm on ubuntu 18.04, I work with Openvino 2020.1.023 , Python* version 3.6.9 and Cmake* version 3.10.2. I already run all the installation steps. I executed the command <INSTALL_DIR>/deployment_tools/demo/demo_speech_recognition.sh, this one launches the execution of two demos. The first offline speech recognition demo works fine but the second live speech recognition demo launches a GUI where I get the following error messages when I click on one of the "Recognize" buttons

In the GUI: Failed to initialize ASR recognizer

In the console: 

./run_demo.sh
[setupvars.sh] OpenVINO environment initialized
target_precision = FP32
Using model configuration file /home/activateur/openvino_models/ir/intel/lspeech_s5_ext/FP32/speech_lib.cfg
Run Inference Engine live speech recognition demo

/usr/local/lib/python3.6/dist-packages/wx/lib/pubsub/__init__.py:25: wxPyDeprecationWarning: wx.lib.pubsub has been deprecated, plese migrate your code to use pypubsub, available on PyPI.
  wx.wxPyDeprecationWarning)
[ INFO ] feature_vector_size: 143
[ INFO ] Using feature transformation /home/activateur/openvino_models/ir/intel/lspeech_s5_ext/FP32/lspeech_s5_ext.feature_transform
[ ERROR ] Could not parse number '-74.24161' in /home/activateur/openvino_models/ir/intel/lspeech_s5_ext/FP32/lspeech_s5_ext.feature_transform
[ ERROR ] Add shift dimension mismatch (0 != 143)
[ERROR] speech_engine.py::initialize | Failed to initialize engine (lib: 'libspeech_library.so'). Status code: -4
[ERROR] speech_proxy.py::speech_process | Failed to initialize ASR recognizer
[ERROR] speech_manager.py::initialize | Failed to initialize Speech Library process
[ERROR] speech_engine.py::initialize | Failed to initialize engine (lib: 'libspeech_library.so'). Status code: -4
[ERROR] speech_proxy.py::speech_process | Failed to initialize ASR recognizer
[ERROR] main_controller.py::_recognize_wave_start | Failed to initialize ASR recognizer

I also tried to execute the python script and the run_demo.sh located in the directory <INSTALL_DIR>/data_processing/audio/speech_recognition/demos/live_speech_recognition_demo where I added the lib file with libspeech_library.so and the models file containing the en-us with speech_lib.cfg however I get the same error.

Thanks for your help