CLASSIFICATION COMPARISON OF DEEP LEARNING MODELS ON AN IMAGINARY SPEECH EEG DATASET
| dc.contributor.author | ONeill, Joseph Thomas | |
| dc.date.accessioned | 2023-05-01T13:05:51Z | |
| dc.date.available | 2023-05-01T13:05:51Z | |
| dc.date.updated | 2023-05-01T13:05:54Z | |
| dc.description.abstract | The restoration and retention of speech in patients with degenerative diseases such as Amyotrophic Lateral Sclerosis (ALS) is a research topic that could benefit many patients worldwide. One path for this is using Brain-Computer Interfaces (BCIs) to classify imaginary speech from electroencephalogram (EEG) data. Currently, this research niche lacks realistic model options and commonly uses computationally expensive preprocessing methods that would not pair well with a commercially viable BCI device. These devices would typically be small, and they would most likely lack significant computational resources. This study aims to eliminate computationally expensive preprocessing and replace it with a simple data transformation and z-score normalization. We also compare the accuracy and performance of five individual deep learning models using a self-recorded, binary, imaginary speech EEG dataset. These models have reportedly high accuracy in their original use-cases. We compare them in this study to find a commercially viable deep learning model for real-world use with a BCI. | |
| dc.identifier.uri | https://repository.library.cofc.edu/handle/123456789/5418 | |
| dc.language.rfc3066 | en | |
| dc.title | CLASSIFICATION COMPARISON OF DEEP LEARNING MODELS ON AN IMAGINARY SPEECH EEG DATASET | |
| dspace.entity.type |