Trying edu.cmu.sphinx.pocketsphinx with processRaw to detect a keyword.
I have setup the SpeechRecognizer's decoder directly with getDecoder().setKeyphrase(KWS_SEARCH,KEYPHRASE); First calling decoder.startUtt(); then few calls to processRaw with slices of a PCM buffer then calling decoder.endUtt(). I get an empty hypothesis for decoder.hyp();
Have tried few different values for setKeywordThreshold. No luck.
What's missing?
To my understanding When calling processRaw the last parameter full_utt should be False until calling endUtt, is that correct?
When should full_utt be set to True? and how does it affect the recognition?
Edited: Need to mention that at first I am trying to detect "oh mighty computer" phrase which is exactly the demo phrase which is already recognized correctly using the SpeechRecognizer's own RecognizerThread yet when using processRaw got no detection. The audio conditions are the same for both attempts.
Thanks.
You missed the tutorial recommendation: