Ditch the Pedometer... NewScientist
"Amit Pande of the University of California, Davis, designed an activity tracking algorithm for smartphones. It works by training a neural network, which functions like a simplified human brain, to recognise features in the data gathered by the phone's accelerometer and barometer. The system also takes account of variables like age, gender, height and weight to estimate energy expenditure.
The team compared their system to the Fitbit and Nike's Fuelband, the leading activity monitors on the market, as well as a professional, wearable calorimeter. It was more accurate than the commercial devices and closely matched the calorimeter.
Fitbit and Fuelband tested particularly badly on stairs. In trial runs up and down four flights of stairs, the commercial devices estimated that more energy was expended going down than up – clearly untrue. In contrast, the smartphone performed better thanks to its barometer, which measures the tiny changes in atmospheric pressure that tell the device when someone is going up and down stairs."
"This is a new market, so in the beginning people are not so worried about accuracy," Pande says. "But we want accurate data so that physicians can use it to improve their understanding of human beings."