For a given channel, we could define the global maximum ` 
' for a
given channel as
Similarly, we could define a distance measure on a single channel as:
These global features can be useful for classification; in fact, in the past this has been one of the ways of doing temporal classification using existing classification tools. For example, with the sign language recognition domain, approximately 30 per cent accuracy can be obtained by looking at the minima and maxima of the x, y and z axes alone [Kad95]. However, they are usually not sufficient for doing high-accuracy classification. In addition, they are usually insufficient for producing meaningful descriptions.
In the Blues and Reds domain, a simple global variable we could use would be the fraction of the time that the channel c is true.