To use the kalman filter to analyze the signals you first have to create a state-space model which will be directly passed with the noise covariance to the kalman function.
Refer the following link to know more about how to create a state-space model.
Refer the following link to know more about how to apply kalman filter using a state space model
Incase you just want to estimate the orientations from the readings of accelerometer and gyroscope readings, you could use the imufilter function which returns a kalman filter object which could be used on the signal readings directly.
Refer to the following link for more informantion about the imufilter function and about the FUSE object which will be returned when imufilter is used.
Refer to the sample code below to know how it is used. In the below example the simpler version is given but you could set different properties for the filter by passing different inputs in the imufilter function.
FUSE = imufilter;
[orientation,angularVelocity] = FUSE(accelReadings,gyroReadings);
Hope this helps!