System and apparatus for the detection of current vehicle class using accelerometer and GPS inputs (16-Nov-2009)

System and apparatus for the detection of current vehicle class using accelerometer and GPS inputs

In some systems the current transport type (car, bike, train, bus, horse, etc.) provides useful contextual information. Depending on this information behaviour of the mobile device, or applications on it, may change.

    For example, a mapping or GPS application needs to display different information to the user depending on whether they were walking (footpaths), cycling (cycle paths, bridleways and roads), or driving (roads).

    Current solutions to this problem involve querying the user; requiring manual entry of their transport type or display preferences.

    This proposal describes the use of components (such as the GPS and accelerometer) on a mobile device to automatically detect the current transport type using a combination of wavelet analysis/matching and speed profiles.

    The device would require a catalogue of transport profiles. This could be created pre-deployment or the system could allow for the definition of additional profiles using either manual entry or sampling of current conditions and measurements. The current wavelet profile can be generated from multiple sources, one such mechanism would be to read the input from an accelerometer.

    Example profiles for: transport mode sample '

Y

' axis (up and down) wavelet

cycle

    wavelet profiles would have characteristics of bumpy (less suspension than cars)

swaying sharper

turns

leaning into

turns
speed: ~ 0 - 25mph

horse

    walk, trot, canter and gallop are all separate wavelet profiles, as horses have different gaits appropriate speed ranges

trot:

pedestrian
walking, running,

ogging etc. are all separate wavelet profiles appropriate speed ranges

j

motor vehicle
wavelet profiles

would have characteristics of smooth (more suspension than bikes)
wide turns fast speed
fast acceleration

The analysis process would follow the method:
1) Take a set of wavelet samples representing short-term historical movement in the three axes; x,y,z.
2) Use wavelet matching techniques to establish a degree of similarity between the profiles and the taken samples.
3) Take measurements from other appropriate sources such as speed, etc.
4) Based on these measurements and wavelet similarities, make a determination...