A Temporal Kernel-Based Model for Tracking Hand Movements from Neural Activities

Subject-specific modulation of local field potential spectral power during brain–machine interface control in primates

“…CLDA is an emerging paradigm for improving or maintaining the online performance of BMIs. By adapting the decoder's parameters during closed-loop BMI operation (i.e., while the subject is using the BMI), CLDA algorithms aim to match the decoder's output to the subject's particular pattern of neural activity [17,[19][20][21][22][23][24][25][26][27]. In our experiments, we used the SmoothBatch CLDA algorithm to adapt the KF decoder's parameters during initial closed-loop control.…”

“…For example, in our experiments, we used Gilja et al's method for inferring intended cursor kinematics ('innovation 1' of the ReFIT-KF algorithm), which assumes that the subject always intends to move to directly toward the current target [17]. Other methods for estimating intended movements could also be used in conjunction with SmoothBatch, such as Shpigelman et al's supervised method [21] or Li et al's unsupervised method [24]. Using the estimate of intended cursor kinematics and the recorded neural activity, SmoothBatch then constructs batch maximum likelihood estimates, Ĉ and Q, of the C and Q matrices using the following equations:…”

Context-aware adaptive spelling in motor imagery BCI

“…Works employing reinforcement learning (RL) and optimal control have been mostly studied in this context [39][40][41][42]. Additionally, since in this field the BCI usually decodes continuous variables (e.g., kinematics), co-adaptation has mainly been studied with respect to adaptive KFs, artificial neural networks, autoregressive models (AR) or custom approaches [43]; the term closed-loop decoder adaptation has been coined for such co-adaptive frameworks [44][45][46][47][48][49].…”

Advantages of closed-loop calibration in intracortical brain–computer interfaces for people with tetraplegia

“…In continuous spike-based BCIs such as the one used in this study, various adaptive decoding methods have been tested in CL neural control in able-bodied non-human primates (Taylor et al 2002, Helms Tillery et al 2003, Wahnoun et al 2006, Jarosiewicz et al 2008, Shpigelman et al 2009, Li et al 2011, Gilja et al 2012, Orsborn et al 2012, Sussillo et al 2012 and in people with tetraplegia (Hochberg et al 2012, Collinger et al 2013. In these studies, the decoder was initialized in OL by mapping neural activity to hand movements, to observed target presentations, or to observed effector movements.…”

“…The decoder was then recalibrated using data acquired during CL neural control. In a subset of these intracortical studies (Helms Tillery et al 2003, Shpigelman et al 2009, Li et al 2011, Gilja et al 2012, Sussillo et al 2012, the quality of neural control using the 'CL decoder' was compared to the quality of neural control using the original OL decoder, and in each case, the CL decoder outperformed the OL decoder. However, none of these studies were designed to distinguish among the multiple possible contributions to the superiority of CL decoders, which could include (1) differences in neural activity during the OL versus CL calibration tasks, (2) increased mental engagement during CL, (3) the inclusion of more calibration data in the CL decoder, and/or (4) the ability of CL decoders to adapt to signal nonstationarities.…”