I am doing a research project on the accuracy of physically activity trackers on iPhone. The purpose is develop an algorithm that can determine that a recorded physical step (steps people take when moving around) are valid steps - meaning they were actually walked by the user and not a dog or put on a machine of some kind that vibrates the phone and tricks it into recording a step.
We know the users: Age, Height, Weight and Sex.
We know the reported number of steps the device has recorded for a given time range (A step record).
Data We Have Access To:
Core Motion sensor data which includes, raw accelerometer, gyroscope , pedometer etc. (A complete list can be found here: [login to view URL]). We can access this data for the same time range as mentioned above.
We DO NOT have access to any GPS or location data.
Develop an algorithm that can reasonably determine if any given step record consists of valid steps using the fixed data pints we know (age, height, weight and sex and number of reported steps) and analyzing it against the sensor data we have access to. Bonus if this can be turned into C# code or method so it can be implemented in a mobile app.