Figure 2。 FDR sampling example, where ‘‘Key’’ indicates the microphone is transmitting。
necessary degrees of freedom in the mathematical model that were discovered during the TDD development process。 The method used to perform the alignment is as follows。 Given a series of CVR radio transmissions, (CVRs, CVRe)i m and FDR Microphone keyings, (aFDRs, aFDRe)j n, where s and e indicate start and end times, the two series are cross joined to create m x n trial
displacements, [Si m,j n], according to
2 aFDR1 3
6 aFDR2 7
The result of the mixed binary, linear program is an offset, S, which is the one S out of all the m x n S9s that maximizes the total area of overlap of the two series of transmissions。 This result is then used in a dynamic programming loop to create the aligned transmissions in a format suitable for the Step 2-Optimization。 The pseudo- code for the loop is shown in Listing 1。
Listing 1
For (each CVR transmission)
。CVR1 CVR2 。。。 CVRi 。6 7
6 。 7
ð1Þ
For (each FDR transmission)
If (CVR transmission length 。 aFDR
4 5
aFDRJ
where CVR and aFDR represent the midpoint of the transmission, s + (e2s)/2。0 and the mathematical operation performed on each cross join is simply, aFDR2CVR。
The array of trial displacements are used in a mixed binary, linear programming model, consisting of m x n binary decision variables, SMatchij (where (i,j) m,n) with one constraint such that one and only one SMatchij can be
1。 The objective function is designed to maximize the overlap of the CVR and FDR transmissions and is expressed as
Max
transmission length)
If (aFDR midpoint is contained within (S + CVRs) and (S + CVRe))
{Output Line Match}
If (aFDR transmission length 。 CVR transmission length)
If ((S + CVR) midpoint is contained within aFDRs and aFDRe)
{Output Line Match} End FDR For
End CVR For
Each output line of the matching step consists of six
numbers, as shown in Listing 2。 The CVR times in Listing 2 are the CVR times provided as input to the step 1 model。 The
FDR outputs are both the adjusted FDR times and the raw FDR times (the FDR times provided as input to the model)。
m SMatchi,j
where the farea is a custom function to calculate the pseudo- area of overlap for each of the conditions expressed in Figure 3。 For example, the pseudo-area of overlap for case 4 from Figure 3 is defined by
Listing 2
CVRs CVRe aFDRs aFDRe rFDRs rFDRe
The dynamic programming step also lists those points that did not find a matching CVR or FDR point。 This can
。 。 。。2
sðlÞ i,j eðkÞ
happen if the CVR analyst did not hear or code a