need help with State estimation using Kalman Filtering or time varying kalman filter ...

This is fNIRS signal basically changes in heomoglobin concentration during a particular stimuli in the motor cortex region of the brain.

So i have states = {'active','rest'}.

and 6 features = slope, variance, peak, skewness,kurtosis, variance.All these features are finding out which classifier is able to predict rest or active.... mostly SVM gave me highest accuracy ~75%. So now this offline analysis using classfication learner built in maltab.

Is it possible to predict the same at real time(ie rest and active) using glm kalman filtering or something?

The above screenshot is from allresults_hbo and i got this by doing a group summary of stats_hbo. Both these files are attached.

[CalebJones]

Star Strider

Any ideas ?