Question

Can you please advise a way of binary data fusion?
Here is a a task:
There are n (n is odd) sources of binary labels (0 | 1). So, every data “frame” contains n labels. The task is to produce a single label per frame based on the fusion of all labels. For example:

S1 0 0 0 1 1 1 0 0 0 1 1 0
S2 0 0 1 1 1 1 1 0 0 1 1 1
S3 0 0 0 0 1 1 1 0 0 0 1 0
--------------------------
   0 0 0 1 1 1 1 0 0 1 1 0

The “major voting” was used in this case: 0 0 0 -> 0; 1 1 0 -> 1 etc.

The major voting could be extended in horizontal direction, so that it’s done over k frames for every i-th frame E.g. for k=3:

F1 round( (0+0+0+0+0+0+0+1+0) / 9) = 0
F2 round( (0+0+0+0+1+0+1+1+0) / 9) = 0
F3 round( (0+1+0+1+1+0+1+1+1) / 9) = 1 # was 0 
F4 round( (1+1+0+1+1+1+1+1+1) / 9) = 1 
.. 

Are there any other fusion schemes that come to your mind?
Thank you!

Answer 1

It looks to me you might be interested in

• The tradeoff between reliability, consistency and availability. Here you can read about it with Amazon’s Dynamo as an example.

• Forward Error Correction