Classify the channel environment

The possible solutions includes:

1) Using Kolmogorov-Smirnov test
--> Try to find the difference D on the CDF of received envelope
--> In our case, we can compute received signal envelope by using RSSI and noise floor over some time period.
--> Two Ds are needed to classify three channel models.
--> Cons: computation complexity; the threshold is a critical parameter
http://www.physics.csbsju.edu/stats/KS-test.html


2) Kurtosis index
--> not only for classification, but also identify the quality of link (seem to reflect SNR)
--> Kurtosis is the degree of peakedness of a distribution.
http://mvpprograms.com/help/mvpstats/distributions/SkewnessKurtosis
http://davidmlane.com/hyperstat/A53638.html
http://www.itl.nist.gov/div898/handbook/eda/section3/eda35b.htm

I believe we can use both Skewness and Kurtosis to classify, even for a fine classification.