accelerometer cannot be summarized in transformer anymore

[sc00n]

In the previous versions (pre 1.0.0), the accelerometer was a multidatum. This was usefull as I could add a transformer that immedieately took the average of all the values (and average of the squared values). Therefore I wouldn't need to safe the countless individual values.

In the new version the accelerometer (e.g. NON_GRAVITATIONAL_ACCELERATION) does not give a multidatum anymore. Therefore, the previous approach doesn't really work anymore. Is there away to avoid saving all individual values and to save only the average, using a transformer or some other way?

[bardram]

Good question - I will think about this.

[bardram]

Now I have been thinking (and even implemented a prototype for buffering of data -- see #338).

But I've come to the conclusion that introducing a buffer of data (similar to the old MultiDatum) is the wrong approach.

Instead, we need to define a new Measure which defines what we want to measure. And then provide a probe for doing this. This is similar to the LightProbe which is a BufferingPeriodicStreamProbe that buffers light data and returns metadata on light (incl. max lux and average lux).

This measure and its probe should be part of a sampling package. You can either create your own sampling package or you can provide me with the code and I can add is as part of the sensor sampling package?

-o-

In general - can you provide some (scientific) description of how to process accelerometer/gyroscope data in a standard way? We could implements this as part of CAMS.

[sc00n]

I created some (actually 4, I wasn't able to figure out how to combine everything for the non master branch) pull requests for a new sensor, the NON_GRAVITATIONAL_ACCELERATION_AVERAGE:

• Update sensors.g.dart #343
• Update sensor_probes.dart #342
• Update sensor_package.dart #341
• Update sensor_data.dart #340

In general it works just like the Ambient Light sensor, where the accelerometer is averaged over a period of time. I summarized it in 7 variables:

    // xm, ym, zm: normal averages (or means)
    // xms, yms, zms: averages of squared values
    // n: number of values included
  double? xm; 
  double? ym;
  double? zm;
  double? xms;
  double? yms;
  double? zms;
  int? n;

using the normal average and average of squared value, the standard deviation can also be computed later:

[bardram]

Now the new AVERAGE_NON_GRAVITATIONAL_ACCELERATION measure
is added as part of carp_mobile_sensing 1.3.3.