I am working on an application which will analyse sports performance data (in this case, second-by-second power output data from training cyclists). I need to be able to look at a time series of power data and identify when the athlete was doing interval training (which may only occur during a subset of the session data -- or not at all), with the definition of "interval training" implying alternating periods of high and low power intensity. And I need to separate those intervals out for analysis. The problem is, the data is not so much noisy as irregular. Here is a picture of a plot....

In this example, only the 3 obvious blocks would qualify as sets containing 8, 7 and 6 intervals, respectively. I have looked at moving averages, and attempted to identify intervals via steep changes in slope. To be clear, I am not asking for a step-by-step solution (although that would be great); just a strategy as to how to achieve this in Mathematica.

Thanks in advance for any pointers.

Brad