I'm not sure this is as easy as it first appears to calculate.
You could integrate the line on a graph showing rate of change in vertical G against time, but as the sensor is rotating you would need to record G in all three dimensions and then use a reading from the gyroscope to calculate the vertical G vector for every point in time. Then you could do your integration. Maybe the device already records enough data already...
Alternatively, have you read the bit in the paper above where they say how they get the "stride angle"? They don't actually measure it. They measure where the foot breaks LED beams just above the treadmill surface. Then they assume a perfect parabolic motion between these points and magic the numbers with a formula. As Runscribe already gives stride length they could copy the paper's bodge method.
Edited to correct: I'm suggesting integrating the wrong graph. The Y-axis would need to be an integral of a graph of G against time. This could then be integrated for a second time. The X axis would be time.