Neural Decoding Using Generative BMI Models - Brain-Machine Interface Engineering - page 166

FIgURE 5.11: Zoomed-in example of probabilistic overlap.

movement segmentations) that produced the best classifications individually. Specifically, we pre-
sent the ratio

( ) i i

m

( )

P O
P O

( |

|

λ
λ

)

for each neural channel in a grayscale gradient format. The darker bands

( ) i ( ) i

( )

r

represent ratios larger than 1 and correspond to a higher probability for the movement class. Lighter
bands represent ratios smaller than 1 and correspond to a higher probability for the rest class. The
conditional probabilities nearly equal to one another show up as gray bands, indicating that clas-
sification for the movement or rest classes is inconclusive.
Overall, Figure 5.10 illustrates that the single channel HMMs can roughly predict movement
and rest segments from the neural recordings. In Figure 5.11 , they even seem tuned to certain parts
of the trajectory like rest-food, food mouth, and mouth-rest. Specifically, we see there are more
white bands P O

λ during rest segments and darker bands P O λ ( | λ ) >> P O ( | ) ( | λ ) >> P O ( | )

r m m r

during movement segments.

FIgURE 5.12: Training and testing with neural data.