As Burmese isn't currently a supported language for Android's built-in cloud-based speech recognition, PocketSphinx may be your best bet. Unfortunately, I don't believe there is an existing language model for Burmese, so you will have to build a custom Burmese language model, which requires a large corpus of text and some linguistic expertise.
CMUSphinx - Building Language Model
...additionally, you'll need to train an acoustic model, which requires a lot of recorded audio. For general dictation (i.e., general purpose speech-to-text that can be used by many different people), the Sphinx project folks recommend you have 50 hours of recordings of 200 different speakers.
CMUSphinx - Training Acoustic Model For CMUSphinx
I've had a good experience using PocketSphinx on Android, but only for a small language model (i.e., a command-and-control type of application), and only in English. Doing so for Burmese is possible, but will take a lot of work to build and train the models mentioned above.