Hi!
I have to use the "Feature Aggregator" box, but it isn't for me clear what types of features are extracted.
I have seen some examples and usually the "Feature Aggregator" box is preceded from "Signal DSP" box (doing the x*x), the "Signal Average" box and another "Signal DSP" box (doing log(1+x) ), but i don't get the meaning.
Thank you in advance!!
Anne
Feature aggrator
Re: Feature aggrator
Hi Anne,
In fact the Feature Aggregator, as its name suggests, doesn't extract anything.
It reads incoming features (you can add as many feature inputs as you want), and when one feature has been received on every input, the box aggregates them in one vector.
Basically, it's just a conversion in another stream format : the feature vector.
This box is always used pior to classification boxes :
- Before the Classifier trainer to produce the feature vectors that "feed" the trainer on each class.
- before the Classifier processor to produce the feature vector processed by the classifier to predict the corresponding class.
Thus the feature aggregator aggregates only... one feature for the classification.
I hope this is clear enough
Laurent-
In fact the Feature Aggregator, as its name suggests, doesn't extract anything.
It reads incoming features (you can add as many feature inputs as you want), and when one feature has been received on every input, the box aggregates them in one vector.
Basically, it's just a conversion in another stream format : the feature vector.
This box is always used pior to classification boxes :
- Before the Classifier trainer to produce the feature vectors that "feed" the trainer on each class.
- before the Classifier processor to produce the feature vector processed by the classifier to predict the corresponding class.
Here you extract one feature : log(1+x²) where x is the incoming (filtered) signal. This is called logarithmic band power.have seen some examples and usually the "Feature Aggregator" box is preceded from "Signal DSP" box (doing the x*x), the "Signal Average" box and another "Signal DSP" box (doing log(1+x) ), but i don't get the meaning.
Thus the feature aggregator aggregates only... one feature for the classification.
I hope this is clear enough
Laurent-
Re: Feature aggrator
Ok,
i begin to understand....
But if I utilize more than 1 channels, how is this managed? They are put in a vector? so if I use more than a feature and more than a channel, it comes out a matrix of features?
Thank you
Anne
i begin to understand....
But if I utilize more than 1 channels, how is this managed? They are put in a vector? so if I use more than a feature and more than a channel, it comes out a matrix of features?
Thank you
Anne