Answers are grounded in this paper and cite their sources.
Objective. A major goal of brain–computer-interface (BCI) technology is to assist disabled people with everyday activities. Although lots of information is available on typical movement procedures, integration of this knowledge is rarely found in motor BCI decoding solutions. Approach. Here, we apply a hidden Markov model (HMM) based approach for continuous decoding of finger movements from electrocorticographic recordings from three human subjects. Information about relative frequencies of consecutive finger movements is included in the decoding routine using so-called bi-gram models. Main results. The presented method achieves accuracies up to 73% for continuous decoding of finger movements. Prior knowledge (PK) incorporation further increases decoding accuracies by up to 12.5% (absolute) in a generic BCI setting and by up to 22% in a more specific, task-related setup. Significance. The results provide evidence for the importance of PK incorporation for motor BCI decoding. We show that this can be done conveniently using HMM decoders. Our results strongly suggest the extension of the use of HMMs from conventional speech-related topics (like spelling devices) towards motor BCI solutions.
No takes yet. Share an insight, caveat, or question.
Pfeiffer et al. (2017) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: