The Next Generation of Slides

NextSlide utilizes CMUSphinx speech recognition software to recognize vocal cues which trigger images and sounds.

When the application starts, users are taken to the main activity, where they can choose to either edit or begin presentations. Upon selecting the edit option, an activity that allows the association of words to actions will begin. A list of current associations is displayed; new ones can be added and existing ones can be deleted. When adding new associations, a word is set and then it is specified if the word will trigger an image or a sound. These images and audio files can be extracted from the device's memory. The associations are stored in a hashmap to allow for easy retrieval of actions when given a key word.

When all associations are configured, the user can begin the presentation from the main activity. When initialized, the presentation will convert the key words hashmap into a file readable by CMUSphinx and then initialize speech recognition. Speech recognition runs in a loop until the activity is destroyed. While it is running, it listens for the key words and triggers the associated action.

The recognizer is streamlined to react as quickly as possible to speech input. Actions are processed within the onPartialResults() function, which is ran continuously during speech recognizer transactions. We ensure that no action is repeated by updating a list of triggered actions whenever a key word is said, which gets reset every transaction. The end result is a dynamic program that accurately and quickly responds to user speech input.