Sections
Introduction
Speed profiles are probability distribution curves that capture movement characteristics of a given activity, or even a given activity filtered by a particular criteria to create a sub-profile. Since speed tends to be strongly influenced by physics of the activity and movement resistance, given a large enough dataset among people it is likely to follow a meaningful normal distribution rather than something like a pareto distribution or having no statistical distributions, and therefore is likely to be more predictive independent of the person.
Some examples where these are useful are in auto-classification of activities from GPS Tracks, as well as in time estimates for planned activities.
'Speed' can apply to movements besides the most commonly considered horizontal/linear. It can also apply to any rate of change, such as elevation gain/loss, or rotation around a curve.
Movement Characteristics
Movement characteristics of particular interest are:
Maximum speed
Average speed
Minimum speed
Standard deviations
These help with typical maximum and minimum speeds, as well as confidence in matching an average speed for, say, Deducing an activity.
Sub-Profile
In addition to an overall profile of an activity, there can be cases where an additional criteria correlates with a different speed profile that can add greater accuracy to activity considerations. A common case is slope. People tend to walk or bicycle faster uphill vs. downhill vs. on flats.
For normal distribution curves, fewer data points are needed for better predictions when other variables (random or otherwise) are excluded. For example, how fast people tend to walk on a trail or boulder hop varies not just by the person, but by the terrain. So a given trail or common cross-country route may have it’s down speed trends, at the most granular level. At a broader level for unknown routes or routes without much recorded data, just comparing across characteristics can be more predictive than just, say, walking speed in general.
Some useful sub-profile criteria are:
Slope (+/-)
Weight carried
Terrain (dirt/paved, trail, cl. 1, cl. 2, cl. 3, etc.)
Pre-determined route with historical data