Pre-processing optimization on sound detector application AudiTion (Android based supporting media for the deaf)

Abstract

In this paper, researchers test several algorithms called by features, as part of pre-processing from AudiTion application. AudiTion is a mobile application that help deaf people to know their environment with visualized the sound into text that describes the sound itself. However, AudiTion has some problems, like delay, the load process, and the accuracy in predict sound. The research will use different media (desktop) to overcome the delay. In addition, the researchers justify the correct processing algorithm, use different database and pre-processing feature. The research focus on optimize the pre-processing section in AudiTion. Start with test some features to find the best accuracy in predict some sound sets as database, then compare the result each other. The best feature which result the highest accuracy is implemented into an application. The application does some further tests with some modification to database to make more improvement. Finally, compare the result again with different sound type with similar composition.