UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The performance enhancement of the LF-ASD brain-computer interface : an Energy normalization transform Yu, Zhou


The brain-computer interface (BCI) has emerged as a potential and radically new mode of communication for users with neuromuscular impairments since it provides a communication channel based on human brain activity as opposed to peripheral nerves and muscles. It is a practical problem to detect user commands among spontaneous Electroencephalograph signals. Low-Frequency Asynchronous Switch Design (LF-ASD) is one of the leading means of addressing this problem. Although the performance of the LF-ASD is encouraging, it is not yet sufficient for real world application. The main goal of this research study is to improve the design of the LF-ASD BCI technique, and then evaluate the performance of this modified design. In this work, the Energy Feature related to Voluntary Movement Related Potential (VMRP) was determined. An energy normalization transform was proposed corresponding to this Energy Feature. A simulation model was set up and EEG data from five able-bodied subjects was applied for offline evaluation. The impact of this normalization transform was evaluated in two studies: the impact of this transform on the low frequency EEG and the impact on the performances of the LF-ASD. By analyzing the experimental results, the characteristics of this normalization filter were determined. In Study 1, it was determined that the proposed normalization transform has two major benefits to the low frequency EEG components (0-4 Hz), which is used by the LF-ASD. First, it can decrease the input scale variance. Consequently, it resulted in more stable feature sets and then a higher successful classification rate. Second, it can increase the separation between the VMRP and idle data. Another side benefit is that the proposed normalization transform can also adjust the input scale automatically. In Study 2, the performances of the LF-ASD with and without this normalization transform were compared. For four out of the five subjects, this transform increased the successful classification rate (True Positive rate with the corresponding False Positive rate at 1%) by 7.7%, 8.3%, 8.5% and 18.9% respectively. By applying an alternative energy normalization transform, the performance increased by 0.4% for the fifth subject. In the future with the parameters of the LF-ASD, especially the codebook in the Feature Classifier, derived from the normalized data, the performance could be further improved. The two studies also showed that, although this transform is non-linear in the broadband (0-64 Hz), it does not distort the features used by the LF-ASD. Therefore, it would not hamper the performance of the LF-ASD. The work concluded with the introduction of potential features related to VMRP in the phase spectrum.

Item Media

Item Citations and Data


For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.