ctriant / gr-phasma Goto Github PK

In the case of signal separation there should be a parameter in order to tune the sensitivity regarding the signal bandwidth. Currently, the block assumes a big guardband in the edges of a detected signal.

The OpenCV Random Forest trainer block currently supports a limited number of possible parameters.
On the other hand, the OpenCV 3.2 documentation stipulates a much larger variety of parameterization choices. Thus, effort is needed in order to:

• Refactor existing parameters so that their datatype conforms to the OpenCV 3.2 library.
• Append missing parameters
• Define appropriate default values for each of them

FFT procedure is used in many different blocks, thus it could be wrapped into a generic function or class.
As an example, some parameters of this wrapper could be:

• Total observed bandwidth
• Actual bandwidth of interest
• Size of FFT
• Window preference (i.e. Blackmann-Harris)

In order to facilitate modularity, featureset generation should be extracted from train and prediction blocks. A separate block could be created that:

• Provides the ability to select featureset, number of samples needed
• Input may be stream or message
• The output should be a message of the format that is expected from train and prediction blocks

Signal separation block already handles multiple simultaneous transmissions.
However, there should be the appropriate enhancement to the ncurses console.