Is It Significant? Guidelines for Reporting BCI Performance

Billinger, Martin; Daly, Ian; Kaiser, Vera; Jin, Jing; Allison, Brendan Z.; Müller-Putz, Gernot; Brunner, Clemens

doi:10.1007/978-3-642-29746-5_17

Cited by 76 publications

(87 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 1 shows the mean values and standard deviations over the crossvalidation folds. The significance level is 59 % with α = 0.05 [16] and 7 out of 9 subjects (s1, s2, s4, s5, s7, s8, s9) reached a significant classification accuracy. The average over subjects with a significant classification accuracy is 67 % ± 16 %.…”

Section: Classificationmentioning

confidence: 99%

EEG-Based Classification of Imagined Arm Trajectories

Ofner

Müller-Putz

2014

Biosystems &Amp; Biorobotics

Self Cite

View full text Add to dashboard Cite

Abstract. A brain-computer interface (BCI) in combination with a neuroprosthesis can be used to restore movement functionalities in paralyzed persons. The BCI detects the movement imagination (MI) and the neuroprosthesis transforms it into a real movement. Today's BCIs can detect the process of MI, but not the actual imagined trajectories of, e.g., the arm. Users' control of a BCI would become more intuitive and natural when the detailed MI, i.e., trajectories, are detected. This is called movement decoding. We made a first attempt to decode MIs, and notably, we did not provoke task depended artefacts like eye movements in our paradigm design. However, that made it necessary to restrict the MIs to movements in two orthogonal planes. We classified the movement plane using a decoding method. For this purpose, we decoded the imagined trajectory and correlated it with two assumed movement trajectories, and then assigned the MI to the assumed movement with the higher correlation. That way, we reached a significant classification accuracy in 7 out of 9 subjects, and showed indirectly the decoding of imagined movements.

show abstract

Section: Classificationmentioning

confidence: 99%

EEG-Based Classification of Imagined Arm Trajectories

Ofner

Müller-Putz

2014

Biosystems &Amp; Biorobotics

Self Cite

View full text Add to dashboard Cite

show abstract

“…This is repeated until every run was in the testing set once. Thus, trials from the same run can never occur in both the training and the testing set, and every trial is used for testing exactly once (see also [34]). The training sets are used for initialization of our procedure, and results are obtained from the respective testing sets.…”

Section: Outer Cross-validationmentioning

confidence: 99%

Single-trial connectivity estimation for classification of motor imagery data

2013

Self Cite

View full text Add to dashboard Cite

Objective. Many brain–computer interfaces (BCIs) use band power (BP) changes in the electroencephalogram to distinguish between different motor imagery (MI) patterns. Most current approaches do not take connectivity of separated brain areas into account. Our objective is to introduce single-trial connectivity features and apply these features to BCI data. Approach. We introduce a procedure for extracting single-trial connectivity estimates from vector autoregressive (VAR) models of independent components in a BCI setting. Main results. In a simulated BCI, we demonstrate that the directed transfer function (DTF) with full-frequency normalization and the direct DTF give classification results similar to BP, while other measures such as the partial directed coherence perform significantly worse. Significance. We show that single-trial MI classification is possible with connectivity measures extracted from VAR models, and that a BCI could potentially utilize such measures.

show abstract

“…It is important to have benchmark criteria which not only provide a quick and easy means of identifying the relative success of the technique but also act as a standard to compare the relative performance of all the techniques. Possible criteria are discussed in (Billinger et al, 2012) and include classification accuracy,…”

Section: Bci Evaluationmentioning

confidence: 99%

Translational Algorithms: The Heart of a Brain Computer Interface

Singh

Daly

2014

Brain-Computer Interfaces

Self Cite

View full text Add to dashboard Cite

Brain computer Interface (BCI) development encapsulates three basic processes: data acquisition, data processing, and device control. Since the start of the millennium the BCI development cycle has undergone a metamorphosis. This is mainly due to the increased popularity of BCI applications in both commercial and research circles. One of the focuses of BCI research is to bridge the gap between laboratory research and commercial applications using this technology.A vast variety of new approaches are being employed for BCI development ranging from novel paradigms, such as simultaneous acquisitions, through to asynchronous BCI control. The strategic usage of computational techniques, comprising the heart of the BCI system, underwrites this vast range of approaches. This chapter discusses these computational strategies and translational techniques including dimensionality reduction, feature extraction, feature selection, and classification techniques.

show abstract

Is It Significant? Guidelines for Reporting BCI Performance

Cited by 76 publications

References 53 publications

EEG-Based Classification of Imagined Arm Trajectories

EEG-Based Classification of Imagined Arm Trajectories

Single-trial connectivity estimation for classification of motor imagery data

Translational Algorithms: The Heart of a Brain Computer Interface

Contact Info

Product

Resources

About