Biomedical Engineering Reference
In-Depth Information
FIgURE 2.13:
Local neuronal correlations time-synchronized with hand position and velocity.
2.9 IMPlICaTIoNS FoR BMI SIgNal PRoCESSINg
One of the most important steps in implementing BMI optimal signal processing techniques for
any application is data analysis and the understanding of neural representations. Here, the reader
should take note that optimality in the signal processing technique is predicated on the matching
between the statistics of the data and the a priori assumptions inherent in any signal processing
technique [
97
]. In the case of BMIs, the statistical properties of the neural recordings and the analy-
sis of neural ensemble data are not fully understood. Hence, this lack of information means that the
neural-motor translation is not guaranteed to be the best possible, even if optimal signal processing
is utilized (because the criterion for optimality may not match the data properties). Despite this re-
ality, through the development of new neuronal data analysis techniques we can improve the match
between neural recordings and BMI design [
16
,
35
]. For this reason, it is important for the reader
to be familiar with the characteristics of neural recordings that would be encountered.
Some interesting issues to consider include the following:
•
How much information is encoded in the neural population response? Noise and the finite
number of neurons sampled can influence how one interprets the representation.
