Chapter 3 Design of Coupling Coding in SBR
3.2 Decision of Shared T/F Grid
(+*
()*
(+*
()*
(+*
, , ' &
()*
(+*
()*
(+*
()*
(+*
, ,
' &
()*
(+*
, , ' &
()*
(+*
, ,
! - " .
&
# &
Figure 12: The syntax of the SBR extension data elements in coupling and non-coupling modes
3.2 Decision of Shared T/F Grid
Instead of using the individual set of time segments as the normal L/R mode, there can be only a common segment set in the coupling mode. Although the sharing of side information can save bits, the quality artifact may occur due to inaccurate segment. Hence, the quality degradation should be considered. For the optimal time segments of the signal subbands of the L/R channels in the normal mode, a decision method based on dynamic programming approach has been proposed in our other work [9]. In this section, the modified decision method for coupling mode is proposed to determine the optimal common segment set and measure the affect for quality.
( )
Figure 13: Diagram of shared time segment in coupling and non-coupling modes
3.2.1 Design of T/F Grid by Dynamic Programming in Non-coupling Mode In [9], a decision method of T/F grid by the dynamic programming (DP) in non-coupling mode has been proposed. The basic concept of the DP method is to search the optimal grid in the all possible grids in individual channel by an efficient recursive procedure. The resultant grid G∗ searched by the method will be an optimal solution to make the average of the energy difference (reconstructed energy error) to the original signal energy ratios (DSR) in all quality measurement units, named critical units, minimum. That is,
( ( ) )
the reconstruction error of the critical unit c in the frame. The lengths of the critical unit are defined as four sample points and the critical band bandwidth for time and frequency direction respectively in [9].The number of the time borders and the associated frequency resolution determine the total number of the girds and also affects the resultant DSR. The dynamic programming for DSR analysis is shown as
{ }
where i, j are the border of the time slot consisting of two samples, k is the number of the time borders, and u is the number of the high resolution envelopes. The notationDik,,ju means the optimal DSR from i to j with k time borders and u high resolution envelope. According to [9], there are ten different bit-consuming stages defined in the dynamic programming method. Each stage indicates the different number of time borders and high resolution envelopes in one SBR frame. The scenarios of the ten stages are described in Table 2. Figure 14 illustrates the optimal partition from i to j with 3 time borders and 2 high resolution envelopes.
Table 2: The scenarios of the ten bit-consuming stage
i j
Figure 14: An example of the optimal partition from i to j time unit with 3 time borders and 2 high resolution envelopes
Figure 15 is the flowchart of the dynamic programming method for searching optimal T/F grid. The loop will consider all passable resolution grids. In the loop, it will have an objective function for determining the optimal T/F grid in the same bit-consuming stage. There is another efficiency checking for switching different stages. The dynamic programming method searches the optimal grid from the lower bit-consuming stages to the higher bit-consuming stages. Because the different bit-consuming stages have different requirement of bits, the grid decision in the different stages must consider the tradeoff of bits and quality.
Stage > 9
Record efficient grid Record efficient grid No
Yes
Yes
No Find optimal T/F grid Find optimal T/F grid
EndEnd
Record efficient grid Record efficient grid No
Yes
Yes
No Find optimal T/F grid Find optimal T/F grid
EndEnd Begin
Begin
Figure 15: Flowchart of the DP method proposed in [9]
3.2.2 Design of T/F Grid by Dynamic Programming in Coupling Mode Because of the sharing of the T/F grid, the two criterions used in the above DP method must be modified to simultaneously consider the content of two channels in the coupling mode. There is an objective function which measures the grids in the DP search method. In the normal mode, the objective function is defined as the DSR value described above. To consider both the two DSR values from L/R channels in the coupling mode, the objective function is modified as
(
0, 1)
adopted in the criterion. The optimal grid is the minimum solution of the objective function (11).On the other hand, the iteration criterion of the DP method in the normal mode involves the improvement of DSR in the current resolution. If the improvement of DSR is over the threshold depending on the bit rate, it will update the higher resolution T/F grid to improve the quality. The improvement is defined as
DSR DSR
E= '− , (12)
where DSR' is the optimal DSR for the preceding bit-consuming stage. Similarly, in the coupling mode, there are two improvements of DSR which are defined as
1
The modified iteration criterion is to satisfied the two conditions as follows
(
0 1)
1 iteration criterion can ensure that the improvement of both the DSR values of the two channels are over a low bound, and at least one of the DSR improvements can exceed the large degree to show the efficiency of the new stage.Table 3: Comparison of the grid criterion in the normal/coupling mode
’ –
’ –
’ –
!
! "
#
! "
"
$
% ##
& % ' #
( % #
’ –
’ –
’ –
!
! "
#
! "
"
$
% ##
& % ' #
( % #
) *) %#
+ % # ' * %