(Georgia Institute of Technology, 2023-01-18)
Gemicioglu, Tan
Amyotrophic Lateral Sclerosis (ALS) is a debilitating movement disability that causes patients to gradually lose their ability to voluntarily control their muscles. In some cases, patients who are "locked-in" are unable to move any muscles, leaving them with no means of communicating with caregivers. Brain-computer interfaces (BCIs) attempt to create a means of communication directly through brain activity, removing the need for movement. BrainBraille is a novel interaction method for BCIs, enabling complex text-based communication using attempted movements with a six-region pseudo-binary encoding. In this dissertation, I explore a wearable BCI using functional near-infrared scanning (fNIRS) to make BrainBraille mobile. In an early study, I show that transitional gestures based on executed movements of two hands can be classified in two participants with up to 93% accuracy. I explore how transitional gestures can benefit BrainBraille by expanding the vocabulary and enabling faster responses. Finally, I evaluate future paths for integrating passive haptic training into BrainBraille to reduce the physical exertion needed to learn a BCI for ALS patients.
(Georgia Institute of Technology, 2022-05)
Bhardwaj, Sukriti
IMUTube generates on-body sensor-based human activity recognition data by leveraging the abundance of videos on platforms like YouTube as a dataset for training human activity recognition models. This thesis deals with the development of a user interface for IMUTube which will allow a user, such as a computer vision researcher or healthcare workers, to generate human pose detection data automatically using IMUTube by inputting custom data while hiding the nuances of the state-of-the-art computer-vision and signal-processing based back-end system. Thus, the user interface creates an interactive web application that makes IMUTube accessible to users from different backgrounds and requires no coding experience.
(Georgia Institute of Technology, 2020-05)
Xiao, Kyle Phillip
BeatMania is a rhythm-action game where players press buttons in response to keysound events to generate music. Rhythm-action game charts (the sequence of keysound events) have traditionally been human authored, since each song level must be creatively organized and correspond an overall pattern or theme. A deep neural network approach is proposed for rhythm-action game chart creation, and a method of level evaluation for co-creative AI is defined. That is, given an arbitrary piece of music, human users can generate BeatMania charts as well as give input to an AI collaborator. The problem is divided into two parts: autonomous chart generation and design interaction. For the chart generation process, a combination of features that include grouping information and audio sample labels are incorporated into an artificial neural network. For the design interaction, principal component analysis is utilized for a proposed reinforcement learning model. The co-creative tool is tested against Markov Chain and LSTM baselines via human trials.