no hypothesis being returned after changing model

[bluemoonj]

hi there,
First time to integrate OpenEars into a new project, we need to dynamic generate models and switch on demand.

OpenEars works fine with first model, it serves our purpose, but once we changed model (our new models are generated with same filename), OpenEars say detected utterance, but never returns hypothesis. I went into the document folder, new model was generated with same filename.

==== some NSLog messages like following.

Pocketsphinx has detected speech.
2013-03-10 00:41:39.984 Globish dictionary[25335:11f03] Pocketsphinx has detected a second of silence, concluding an utterance.
2013-03-10 00:41:40.003 Globish dictionary[25335:11f03] Pocketsphinx is now listening.
2013-03-10 00:41:40.806 Globish dictionary[25335:11f03] Pocketsphinx has detected speech.
2013-03-10 00:41:47.190 Globish dictionary[25335:11f03] Pocketsphinx has detected a second of silence, concluding an utterance.
2013-03-10 00:41:47.224 Globish dictionary[25335:11f03] Pocketsphinx is now listening.

anyone has idea what happened? or is is because I cannot use same filename for new models? please advice.

thanks in advance.

Paul

[Halle Winkler]

Welcome Paul,

Please check out the troubleshooting starting points in this post:

https://www.politepix.com/forums/topic/install-issues-and-their-solutions/

[bluemoonj]

hi Hale, thanks for pointing to the troubleshooting post.
I have followed the post and turn on logging. Some messages in console as follow:

2013-03-10 10:26:49.996 Globish dictionary[4371:907] Pocketsphinx is now using the following language model:
/var/mobile/Applications/7B4BE7D5-073F-454B-9FAC-6EDB1117CF8C/Documents/randomWordsDynamicModel.DMP and the following dictionary: /var/mobile/Applications/7B4BE7D5-073F-454B-9FAC-6EDB1117CF8C/Documents/randomWordsDynamicModel.dic
2013-03-10 10:26:51.587 Globish dictionary[4371:3a03] Speech detected…
2013-03-10 10:26:51.591 Globish dictionary[4371:907] Pocketsphinx has detected speech.
2013-03-10 10:26:52.756 Globish dictionary[4371:3a03] Stopping audio unit.
2013-03-10 10:26:52.760 Globish dictionary[4371:907] Pocketsphinx has detected a second of silence, concluding an utterance.
2013-03-10 10:26:52.891 Globish dictionary[4371:3a03] Audio Output Unit stopped, cleaning up variable states.
2013-03-10 10:26:52.899 Globish dictionary[4371:3a03] Processing speech, please wait…
INFO: cmn_prior.c(121): cmn_prior_update: from
INFO: cmn_prior.c(139): cmn_prior_update: to
INFO: ngram_search_fwdtree.c(1549): 632 words recognized (9/fr)
INFO: ngram_search_fwdtree.c(1551): 1737 senones evaluated (25/fr)
INFO: ngram_search_fwdtree.c(1553): 691 channels searched (10/fr), 0 1st, 691 last
INFO: ngram_search_fwdtree.c(1557): 691 words for which last channels evaluated (10/fr)
INFO: ngram_search_fwdtree.c(1560): 0 candidate words for entering last phone (0/fr)
INFO: ngram_search_fwdtree.c(1562): fwdtree 0.15 CPU 0.210 xRT
INFO: ngram_search_fwdtree.c(1565): fwdtree 1.38 wall 1.998 xRT
INFO: ngram_search_fwdflat.c(316): Utterance vocabulary contains 2 words
INFO: ngram_search_fwdflat.c(951): 85 words recognized (1/fr)
INFO: ngram_search_fwdflat.c(953): 204 senones evaluated (3/fr)
INFO: ngram_search_fwdflat.c(955): 131 channels searched (1/fr)
INFO: ngram_search_fwdflat.c(957): 131 words searched (1/fr)
INFO: ngram_search_fwdflat.c(959): 59 word transitions (0/fr)
INFO: ngram_search_fwdflat.c(962): fwdflat 0.03 CPU 0.043 xRT
INFO: ngram_search_fwdflat.c(965): fwdflat 0.05 wall 0.076 xRT
INFO: ngram_search.c(1217): not found in last frame, using .67 instead
INFO: ngram_search.c(1269): lattice start node .0 end node .59
INFO: ngram_search.c(1297): Eliminated 0 nodes before end node
INFO: ngram_search.c(1402): Lattice has 17 nodes, 2 links
INFO: ps_lattice.c(1367): Normalizer P(O) = alpha(:59:67) = -2846433
INFO: ps_lattice.c(1405): Joint P(O,S) = -2846433 P(S|O) = 0
INFO: ngram_search.c(891): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(894): bestpath 0.00 wall 0.001 xRT
2013-03-10 10:26:52.963 Globish dictionary[4371:3a03] Pocketsphinx heard “” with a score of (0) and an utterance ID of 000000002.
2013-03-10 10:26:52.971 Globish dictionary[4371:3a03] Hypothesis was null so we aren’t returning it. If you want null hypotheses to also be returned, set PocketsphinxController’s property returnNullHypotheses to TRUE before starting PocketsphinxController.
2013-03-10 10:26:52.975 Globish dictionary[4371:3a03] Checking and resetting all audio session settings.
2013-03-10 10:26:52.979 Globish dictionary[4371:3a03] audioCategory is correct, we will leave it as it is.
2013-03-10 10:26:52.985 Globish dictionary[4371:3a03] bluetoothInput is correct, we will leave it as it is.
2013-03-10 10:26:53.021 Globish dictionary[4371:3a03] categoryDefaultToSpeaker is correct, we will leave it as it is.
2013-03-10 10:26:53.028 Globish dictionary[4371:3a03] preferredBufferSize is correct, we will leave it as it is.
2013-03-10 10:26:53.037 Globish dictionary[4371:3a03] preferredSampleRateCheck is correct, we will leave it as it is.
2013-03-10 10:26:53.043 Globish dictionary[4371:3a03] Setting the variables for the device and starting it.
2013-03-10 10:26:53.052 Globish dictionary[4371:3a03] Looping through ringbuffer sections and pre-allocating them.
2013-03-10 10:26:53.138 Globish dictionary[4371:3a03] Started audio output unit.
2013-03-10 10:26:53.146 Globish dictionary[4371:3a03] Listening.
2013-03-10 10:26:53.151 Globish dictionary[4371:907] Pocketsphinx is now listening.
2013-03-10 10:26:56.420 Globish dictionary[4371:3a03] Speech detected…
2013-03-10 10:26:56.427 Globish dictionary[4371:907] Pocketsphinx has detected speech.
2013-03-10 10:26:58.005 Globish dictionary[4371:3a03] Stopping audio unit.
2013-03-10 10:26:58.009 Globish dictionary[4371:907] Pocketsphinx has detected a second of silence, concluding an utterance.
2013-03-10 10:26:58.137 Globish dictionary[4371:3a03] Audio Output Unit stopped, cleaning up variable states.
2013-03-10 10:26:58.145 Globish dictionary[4371:3a03] Processing speech, please wait…
INFO: cmn_prior.c(121): cmn_prior_update: from
INFO: cmn_prior.c(139): cmn_prior_update: to
INFO: ngram_search_fwdtree.c(1549): 980 words recognized (9/fr)
INFO: ngram_search_fwdtree.c(1551): 2763 senones evaluated (26/fr)
INFO: ngram_search_fwdtree.c(1553): 1070 channels searched (10/fr), 0 1st, 1070 last
INFO: ngram_search_fwdtree.c(1557): 1070 words for which last channels evaluated (10/fr)
INFO: ngram_search_fwdtree.c(1560): 0 candidate words for entering last phone (0/fr)
INFO: ngram_search_fwdtree.c(1562): fwdtree 0.22 CPU 0.209 xRT
INFO: ngram_search_fwdtree.c(1565): fwdtree 1.78 wall 1.681 xRT
INFO: ngram_search_fwdflat.c(316): Utterance vocabulary contains 2 words
INFO: ngram_search_fwdflat.c(951): 127 words recognized (1/fr)
INFO: ngram_search_fwdflat.c(953): 315 senones evaluated (3/fr)
INFO: ngram_search_fwdflat.c(955): 166 channels searched (1/fr)
INFO: ngram_search_fwdflat.c(957): 166 words searched (1/fr)
INFO: ngram_search_fwdflat.c(959): 76 word transitions (0/fr)
INFO: ngram_search_fwdflat.c(962): fwdflat 0.04 CPU 0.034 xRT
INFO: ngram_search_fwdflat.c(965): fwdflat 0.07 wall 0.064 xRT
INFO: ngram_search.c(1217): not found in last frame, using .104 instead
INFO: ngram_search.c(1269): lattice start node .0 end node .2
INFO: ngram_search.c(1297): Eliminated 20 nodes before end node
INFO: ngram_search.c(1402): Lattice has 22 nodes, 1 links
INFO: ps_lattice.c(1367): Normalizer P(O) = alpha(:2:104) = -3090709
INFO: ps_lattice.c(1405): Joint P(O,S) = -3090709 P(S|O) = 0
INFO: ngram_search.c(891): bestpath 0.00 CPU 0.000 xRT
INFO: ngram_search.c(894): bestpath 0.00 wall 0.001 xRT
2013-03-10 10:26:58.227 Globish dictionary[4371:3a03] Pocketsphinx heard “” with a score of (0) and an utterance ID of 000000003.
2013-03-10 10:26:58.234 Globish dictionary[4371:3a03] Hypothesis was null so we aren’t returning it. If you want null hypotheses to also be returned, set PocketsphinxController’s property returnNullHypotheses to TRUE before starting PocketsphinxController.
2013-03-10 10:26:58.238 Globish dictionary[4371:3a03] Checking and resetting all audio session settings.
2013-03-10 10:26:58.243 Globish dictionary[4371:3a03] audioCategory is correct, we will leave it as it is.
2013-03-10 10:26:58.248 Globish dictionary[4371:3a03] bluetoothInput is correct, we will leave it as it is.
2013-03-10 10:26:58.253 Globish dictionary[4371:3a03] categoryDefaultToSpeaker is correct, we will leave it as it is.
2013-03-10 10:26:58.259 Globish dictionary[4371:3a03] preferredBufferSize is correct, we will leave it as it is.
2013-03-10 10:26:58.279 Globish dictionary[4371:3a03] preferredSampleRateCheck is correct, we will leave it as it is.
2013-03-10 10:26:58.296 Globish dictionary[4371:3a03] Setting the variables for the device and starting it.
2013-03-10 10:26:58.300 Globish dictionary[4371:3a03] Looping through ringbuffer sections and pre-allocating them.
2013-03-10 10:26:58.389 Globish dictionary[4371:3a03] Started audio output unit.
2013-03-10 10:26:58.394 Globish dictionary[4371:3a03] Listening.
2013-03-10 10:26:58.398 Globish dictionary[4371:907] Pocketsphinx is now listening.
2013-03-10 10:26:59.336 Globish dictionary[4371:907] Pocketsphinx has suspended recognition.

anyone has idea why there is no hypothesis being returned? It is strange that the first round before changing model always work, but just after changing model, no more hypothesis…
any help is appreciated.

Paul

[Halle Winkler]

OK, please generate the new models with a unique filename instead of the same one as the previous model. I expect this will fix your issue.

[bluemoonj]

hi Halle,

thank you for the help. indeed the filename seems cannot be the same as current using model.

so my solution is to use two different filenames and just cycle through them, it works fine so far.

again, your help is much appreciated!

Paul

[Halle Winkler]

I’m happy to hear it helped!

[Author]

Posts