Pocketsphinx VS Google Recognition API

Question

I am creating an Android speech recognition application that will act as a virtual personal assistant with tasks such as:

Make appointments / reminders

Weather Info

General Questions Tungsten | Alpha / Wikipedia - (i.e. who led the Ghostbusters, what is the £ - $ exchange rate)

My question is using Pocketsphinx or Google APIs?

I originally installed this using "android.speech.RecognitionListener", it worked fine, however I want to implement a keyword definition so that the user does not need to interact other than just speak.

Apparently, the Google API does not support this, so I considered using pocketsphinx for this and still use google for the rest of the application (as I heard pocketsphinx is not so accurate?) However, they do not get along with each other, since both cannot take a microphone.

Is there a good way to switch between recognizers? (maybe even import both into the same project)

Should I just go with my pockets and deal with lower accuracy?

Suggestions would be helpful

Greetings

Answer 1

For those who want to implement a similar project, I found a job. It is abit hacked and not quite clean, but it works.

Using a speech recognizer for Android with an on / off switch, as in many examples on the Internet, when onResults returns, the string will be checked for the specified "hot word", if not, drop the string, if so, handle it. After the request has been processed and the text responds, open the switch button programmatically, ensuring constant listening.

Do the same on "onError". I also had this onPartialResults, but it looks like it made a stream of crashes, not quite sure why, but once it was deleted, everything worked fine.