Offline Dependent Speech Recognition for Android

When collecting extensive information through a data collection application for an Android device, an easier input method such as voice input becomes an attractive option.  Dependent speech recognition software provides tools necessary to understand the speaker's natural speech.  This type of speech recognition is necessary to allow the speech to text option.  However, most speech recognition software for mobile devices are limited by the online component needed.  Notwithstanding, there are situations where an offline approach to speech recognition is required. This is the case when civil engineers or geologists survey terrain in the field. Thus, a need for a comprehensive understanding of possible offline software solutions versus online software solutions is paramount prior to an appropriate selection.  This research will outline the potentials and possible limitations incurred when the online server side processing is reduced to a mobile device processing.  These limitations are expressed in terms of, but not limited to, grammar size, acoustic models, vocabulary and language.  The use of open source speech recognition offline software solutions, such as PocketSphinx, to achieve the speech to text option are the focus of this research.