"P300Classifier: doubt with the number of features"

[lauramartingc]

Hi,
we're using the P300Classifier tool to create a classifier for P300 online experiments. We set the following parameters:

MaxModelFeatures = 60
Penter = 0.1000
Premove = 0.1500
Spatial filter = CAR
Decimation frequency (Hz) = 20
Channel set = 1 2 3 4 5 6 8
Response window (ms) = 0 800

We generate the feature weights and save them. When we load them in the online experiment, the created classifier has 780 rows. If we set the maximum number of features to 60, ¿why the classifier has 780?

Many thanks in advance.

Laura

[gschalk]

Because the classifier online only has access to the signal at the sampling frequency. Thus, for each feature, it essentially uses the same weight <downsamplingfactor> times, which averages, and thus downsamples, the signal down to the frequency that you provide (e.g., 20 Hz).

Gerv

[lauramartingc]

Thank you so much Gerv.