ESP
ESP copied to clipboard
damellis
The Example-based Sensor Predictions (ESP) system applies machine learning to real-time sensor data.
That allows pausing the use of some other FeatureExtraction module. The specific other FeatureExtraction module used could be specified in a sub-class.
This requires setting up Travis #201 (or other similar services). This [gist](https://gist.github.com/domenic/ec8b0fc8ab45f39403dd) might be helpful.
This would make it easier, for instance, to do audio calibration in the feature domain (after FFT and FFT feature extraction) rather than the time domain.
We could model our current normalizers as calibrators that have no calibration processes / samples. We could even overload useCalibrator() to accept a function and create the Calibrator object internally....
As in normalization, we may want to increase or reduce the number of dimensions.
This isn't high priority, but it might be nice for the user.h examples to provide a name for the pipeline, which then gets displayed in the GUI. Otherwise, there's not...
Depending on the user's setup, we may want to modify the training data supplied by the example author. For instance, if the user is using an accelerometer with a smaller...
This is needed for #66. Right now, some pipeline features do use time, but it's based on the system clock, not on time data associated with the input data. That...