Skip to Content

Sponsors

No results

Tags

No results

Types

No results

Search Results

Events

No results
Search events using: keywords, sponsors, locations or event type
When / Where
All occurrences of this event have passed.
This listing is displayed for historical purposes.

Presented By: Biomedical Engineering

BME PhD Defense: Abdulrahman W. Aref

Improving Brain-Computer Interface Performance By Using Dynamic Methods Based on Analysis of Cognitive State

BME-EVENT Placeholder BME-EVENT Placeholder
BME-EVENT Placeholder
Communication for individuals with severe motor and speech impairments can be very difficult and they find the need for the assistance of augmentative and alternative communication (AAC) systems. Common commercialized AAC systems require some amount of voluntary control and are unusable by individuals with disabilities. Non-invasive brain-computer interfaces (BCIs) are an emerging means of communication for people with severe motor and speech impairments. BCIs allow the user to make selections on the computer just using their brain signals, electroencephalogram (EEG). However, although they are revolutionary for individuals that cannot control other available AAC systems, BCIs have several limitations. Two major limitations of BCIs are: 1) BCIs are static/synchronous in nature; 2) BCIs are susceptible to changes in user attention. Since populations that need BCI technology the most (e.g. amyotrophic lateral sclerosis (ALS)) experience attention impairments, incorporating attention-monitoring features into the BCI would improve BCI performance by reducing errors in these populations. This research presents two dynamic methods developed to help the BCI become more user-aware and allow users to control the BCI at their own pace. Using a well-established negative correlation between alpha band power in the EEG and attention, the first method used alpha band analysis to detect losses in user attention and abstained selections that were unattended to reduce potential errors. The second method, called P300-Certainty, abstained selections that do not reach a specified confidence level. To test both methods, off-line analysis was performed on recorded EEG from 30 subjects using the BCI for spelling. Subjects selected 9 sentences and at least 23 characters per sentence with additional corrections. Alpha band analysis and P300-Certainty were tested off-line, separately and together, on this dataset to determine their efficacy at increasing BCI accuracy by abstaining potential errors. In addition, P300-Certainty was implemented in a BCI-facilitated cognitive assessment to reduce potential errors, as well as, only choosing selections when they reach a specified confidence level. The on-line performance of P300-Certainty was calculated from this data. Alpha band analysis was performed off-line on this on-line data to determine its efficacy at increasing P300-Certainty on-line BCI accuracy.

Alpha band power was shown to be significant between correct and incorrect character selections with a significance of p = 0.01004. Using this significance, alpha band analysis was used to classify selections as correct or incorrect based on the EEG, however it was only useful for accuracy for a subset of subjects (subjects exhibiting high alpha variance). Off-line analysis of P300-Certainty was shown to increase accuracy from 82.01±12.59% to 88.82±8.85% by abstaining potential errors, with a statistical significance of p = 0.038. Furthermore, P300-Certainty and alpha band analysis used together, improved BCI accuracy, over all subjects, more than either method did alone. The increase was statistically significant (p = 0.008) when compared to the raw BCI accuracy. The on-line accuracy of P300-Certainty was 83.62 ± 9.14%.

Alpha band analysis and P300-Certainty abstain potential errors using different, yet orthogonal, methods of measuring attention. Each method abstains potential errors that the other may have not detected. In conclusion, this research has introduced two methods that quantify attention in orthogonal ways that increase BCI accuracy by abstaining potential errors more than either method alone. Using these methods together allows the BCI to be more user-aware and allows the user to type at their own pace.
BME-EVENT Placeholder BME-EVENT Placeholder
BME-EVENT Placeholder

Explore Similar Events

  •  Loading Similar Events...

Back to Main Content