TY  - GEN
AB  - Brain–computer interfaces (BCIs) enable communication and device control without muscle movement and offer important benefits for people with severe motor impairments. This thesis develops a more efficient EEG‑based BCI that detects event‑related potentials (ERPs) from single trials, improving throughput for visual image search. Using cross‑session training, a hybrid generative–discriminative classifier, and computationally efficient learning methods, the system achieves strong ERP detection without performance loss. The thesis also identifies a neural correlate of visual perception by linking ERP responses to task and target difficulty, suggesting that the brain dynamically increases cognitive resource allocation during challenging visual‑recognition tasks.
AD  - Oregon Health and Science University
AU  - Huang, Yonghong
DA  - 2010
DO  - 10.6083/M48P5XGJ
DO  - DOI
ED  - Pavel, Misha
ED  - Erdogmus, Deniz
ED  - Advisor
ID  - 376
KW  - Bioengineering
KW  - Electroencephalography
KW  - Brain
KW  - Neurophysiology
KW  - Brain-Computer Interfaces
KW  - Support Vector Machine
KW  - Evoked Potentials
KW  - Electrophysiology
KW  - Evoked Potentials, Visual
KW  - Neural Networks, Computer
KW  - electric properties
KW  - fisher kernel
KW  - mixed models
KW  - incremental learning
KW  - computational neuroscience
KW  - visual information system
KW  - single-trial erp detection
KW  - pattern recognition
L1  - https://digitalcollections.ohsu.edu/record/376/files/377_etd.pdf
L2  - https://digitalcollections.ohsu.edu/record/376/files/377_etd.pdf
L4  - https://digitalcollections.ohsu.edu/record/376/files/377_etd.pdf
LK  - https://digitalcollections.ohsu.edu/record/376/files/377_etd.pdf
N2  - Brain–computer interfaces (BCIs) enable communication and device control without muscle movement and offer important benefits for people with severe motor impairments. This thesis develops a more efficient EEG‑based BCI that detects event‑related potentials (ERPs) from single trials, improving throughput for visual image search. Using cross‑session training, a hybrid generative–discriminative classifier, and computationally efficient learning methods, the system achieves strong ERP detection without performance loss. The thesis also identifies a neural correlate of visual perception by linking ERP responses to task and target difficulty, suggesting that the brain dynamically increases cognitive resource allocation during challenging visual‑recognition tasks.
PB  - Oregon Health and Science University
PY  - 2010
T1  - Event-related potentials in electroencephalography characteristics and single-trial detection for rapid object search
TI  - Event-related potentials in electroencephalography characteristics and single-trial detection for rapid object search
UR  - https://digitalcollections.ohsu.edu/record/376/files/377_etd.pdf
Y1  - 2010
ER  -