~\
2~~'~'~~~
Process for Simultaneous _Transrnission of Signals from
N Signal Sources
D a s c r p t i a n
Technical Field
The present invention relates to a process fo simultaneous
transmission of signals from N signal sources via a
corresponding number of transmission channels.
State of the Art
Processes in which the individual ('time) signals are
divided into blocks and these blocks are transformed by
transformation or filtering into spectral coefficients
which for their part undergo a data reduction process or
which are coded according to a data reduction re-
spectively, are known. In this connection, reference is
made to, by way of illustration, the overview article
"Perceptual Audio coding" by Jorg FIoupert in Studio-Te-
chnik or the article '°Daten-Diat, Datenreduktion bei digi-
talisierten Audio-Signalen" by Stefanie Renner in Elrad,
1991. These overview articles as well as the PCT ATdo-
cument W0..88/01811 is explicitly referred to with regard
to the explanation of any terms and process steps nat made
more apparent herein.
In a number of cases, it is necessary 'to transmit signals
from several signal sources simultaneously via a
corresponding number of transmission channels. The
transmission of stereo signals via 'two 'transmission
channels is mentioned as the most simple example 'therefor.
. y, . ',~-: ' . . ~: ',
'
~
;,:; . , :: , ', ' ~ . ~:
.' , ; . : ., ', . . '.,v : ' ,,.,..
: . ' '.. ~ ,.
. .;.,~. . ~-',... ; : ~.: : .:;
" ': '~: ,.. ~;
';::
,.. , ,. : ..
. ...:.. . . , : ~ ~ ,. _ ~ . ., . .
;.. .. ', :; .. ,, , ...
~ , ~ . ;; .. ~:. . ~'... ~. .
. ; .;v ~ _ ..
_2_
2
The transmission of signals from N signal sources via a
corresponding number of transmission channels presents the
problem of dimensioning the transmission channels:
If each individual transmission channel is dimensioned in
such a manner that it transmits the "maximum incident bit
flow" (German: Bit-Strom), comparatively large transmis-
sion capacity remains unused "on the average".
Tn the transmission of signals from numerous signal
sources via a corresponding number of transmission
channels, it is known from digital telephone technology
to design the transmission channels for only "average
demand'° and to balance short-term increased demand on
individual channels by allotment from other channels. The
allotment ensues exclusively via signal statistics.
for the state of the art, reference is made to the
following literary sources "Ein digitales
Sprachinterpolationsverfahren mit pradiktionsgesteuerter
Wortaufteilung" by Dr. H. Gerhauser (1950), "Ein digitales
Sprachinterpolationsverfahren mit momentaner
Prioritatszuteilung", by R. Woitowitc 0977) and "Ein
digitales Sprachinterpolationsverfahren mit blockweiser
Prioritatszuteilung" by G.G. Klahnenbucher (1975).
An element of the present invention is that it was
understood that the usual processes in digital telephone
technology for balancing fluctuating demand'in the
transmission of numerous signals via a corresponding
number of transmission channels does not have good results
if the digital signals to be transmitted previously under-
went data reduction, by way of illustration, according to
the so-called OCF process.
-3-
2~.2>~~~
Description of the Invention
'.Che object of the present invention is to provide a pro-
cess for simultaneous transmission of signals from N
signal sources via a corresponding number of transmission
channels with which "data-reduced signals" can be trans-
mitted via transmission channels that are only dimensioned
for "average demand" without any perceptable, i..e. by way
of illustration audible loss in signal capacity.
A solution to this object in accordance with the present
invention is set forth in claim 1. Further improvements of
the invention are the subject matter of the subclaims.
The present invention is based on the fundamental idea
not to conduct the allotment to the individual signals
according to statistical considerations in balancing the
fluctuating requirements during simultaneous transmission
of signals from N signal sources via a corresponding
number of transmission channels, but rather already to
balance the fluctuating demand by appropriate means in the
process step in which the signals are coded for data
reduction.
This inventive fundamental idea is explained in the
following using a preferred embodiment with reference to
the accompanying drawings, showing:
Fig. 1 a block diagram to explain the invented process,
and
Fig. 2a and 2b the invented signal build-up (German:
Signalaufbau)
,
' ~~~~~~ ~ ,
In the invented process the individual signals are divided
into blocks and the blocks are transformed into spectral
coefficients by transformation or filtering.' To balance
the fluctuating demand, the blocks belonging to the
individual signals are divided into sections, and the
respective current sections of all signals are processed
simultaneously. This is graphically illustrated in Fig. 1
by the corresponding "function blocks".
Employing a perception-specific model which, by way of
illustration, when transmitting audio-signals can be a
psycho-acoustic model, the permissable interference is
determined for each section and from it is calculated
the request for the currently required overall transmis-
sion capacity. This calculation of the overall transmis-
sion capacity, d.h. the required number of bits, occurs in ,
all the blocks simultaneously. From all the transmission
capacity at disposal and the currently required overall
transmission capacity, the allotment of maximum transmis-
sion capacity at disposal is calculated for each indivi- ..
dual signal. With each "number of bits" alloted to each
signal, the coding of the individual signal occurs and
accordingly the transmission of this individual signal. In
the simplest case, balancing or equalizing can only occur
in the re-spective required transmission capacity between
the channels.
In the further improvement described in claim 2, there is
a transmission capacity reserve, a so-called bit
reservoir, from which, in the event that the required
overall transmission exceeds the transmission capacity on
the average at disposal, an allotment of transmission
capacity accurs.
-5-.
This bit reservoir is filled whenever the requested trans-
mission capacity is less than the transmission capacity at
disposal (claim 3).
In any case, it is necessary, in order to pr.,event an
increase in the bit reservoir beefing to great, if the
transmission capacity is much smaller than the transmis-
sion capacity at disposal, that there is a forced allot-
ment of bits to the individual channels (claim ~). This
forced allotment occurs preferably only to the channels,
respectively the signal sources, that have reported a need
that is greater than average demand. A substantially grew-
ter demand than the average demand, noteably, means that
these signals are substantially more difficult to code
than usual signals.
In any event, it is preferable according to claim 9 if an
averall block is formed from all the separately coded
signals from the signal sources. This overall block is
composed of a fixed zegion containing information frorn
which the separation of signals can be determined and of
several regions of more flexible length which receives the
coded signals. This is diagrammatically shown in Fig. 2a.
Further saving in transmission capacity is achieved in
that identical input signals are recognized and are
transmitted only once by a suited transmission format
(claim ~). This is diagramatically shown in Fig. 2b.
In any case, the currently required transmission capacity
can be accurately determined or only estimated (claims 7
and 8 ) .
Moreover, in a great extend the invented process can be
conducted parallel. To do this, it is preferable if ac-
cording to claim 10 coding of the indivj.dual signals
already occurs during calculation of the allotment of the
trans-mission capacity for each signal.
Another preferred realization of the invented fundamental
idea is set forth in claim 11:
If the required transmission capacity exceeds the
transmission capacity at disposal and no allotment from
the bit reservoir can occur, the value of the permissible
interference for all the signals can be raised in such a
manner that the required overall transmission capacity
does not exceed the transmission capacity at disposal
(claim 11) .
In the following a numerical example of a manner of
proceeding for audio signals is given. It is explicitly
pointed out that the invented fundamewtal idea is not
restricted to audio signals, but rather also video signals
or other signals underlying a perception-specific
assessment can be treated similarly.
Example of a possible manner of proceeding for audio
signals:
Assuming that y(t) are the sampling values of 'the audio
signal.
1) The audio signal y is broken down or separated ire a
known manner into the sampling values (y(t)), which are
digitized. The digitized sampling values are broken down
ar decomposed into blocks of the length 2n, which in the
N
selected embodiment are overlapping blocks having an over-
lapping of n:
x(k,b) - y(b*n+k) for k=o..2n (b block number).
2) Each block of length n is transformed into spectral
coefficients by transformation, by way of illustration Fast ..
Fourier Transformation or a cosinus transformation:
x(j,b) - SUM(1=0..2n; x(1,b)*f(1)*cos(pi*(21+1+n)
(2j+1) / (4n)) ) fox j=O..n with f(1) - sqrb(2)*sin
(pi * (1+0.5) / (2n) )
3) Each of the blocks is divided into sections and the
energy density is calculated for each section:
E(i,b = ( SUM(k=a(i)+l..a(i-~1); X(k,n)2 ) ) /
(a(i+1)-a(i)) for i=1 ... c,
with the coefficients a(i) being taken from the following
Table 1.
4) The permissible interference is calculated for each
section with a suited psycho-acoustical model, for which
the literature is to be referred. The masking between the
bands is yielded from the permitted interference
T(i,b) - MAX(k=1 ... i-1; E(k,b)*z(i-k) )
the masking in the band:
s(i,b) = max ( E(i,b) * e(i) , T(i,b) )
and the masking between the blocks:
_ _
8 _
ss(i,b) = max ( s(i,b-1)/16 , s(i,b) )
thereon follows the calculation of the required number of
bits for each block.
5) Calculation of the required number of bits for the block:
a) for a coding like in the case of OCF (Huffman coding):
p = p0 + SUM(i=1..C; (a(i+1)-a(i) *(s(i,b)/ss(i,b) ) )
b) for PCM coding (SNR = 6dB/bit)
A scaling factor and the number of bits per sampling value
as additional information are transmitted for each section
p = p0 + SUM(i=l..ct (a(i+1)) * 10/6 * log( E(i,b) /
ss(i,b) ) )
The pertinent values for the individual values, respec-
tively for the individual constants, are given in the form
of tables in the following:
n = 512
c=23 ,;,:;.
p0 = 1200 for OCF (average number of bits per block)
p0 = 345 for PCM (scaling factors: 10 bit/section, coding
of trze number of quantization steps: 5 bits/
section
TRBLE 1
i 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
a(i) 0 4 8 12 16 20 24 28 32 36 40 46 52 60 70 82
--. _g_
i 17 18 19 20 21 22 23 24
a(i) 96 114 136 164 198 240 296 372
TABLE 2
i 1 2 3 4 5 6 7 8 9
e(i) 1e-1 1e-2 1e-3 le--4 1e-5 1e-6 1e-7 1e-8 1e-9 a(1)
0 for i>9
TABLE 3
i 1 2 3 4 5 6 7
e(i) 0.0004 0.0004 0.000 4 0.0004 0.0004 0.0004 0.0004
i 8 9 10 11 12 13 ~ 14 15
e(i) 0.002 0.002 0.00 2 0.004 0.01 0.015 0.025 0.04
i 16 17 18 19 20 21 22 23
e(i) 0.06 0.06 0.06 0.08 0.08 0.11 0.14 0.18
This is followed by the allotment of the number of bits to
the individual signals. F'or this it is assumed that k(k)-
bits are requested for c oding the K input signals while
. psoll number of bits are at disposal.
psum = Sum((p)k))
Now it is necessary to d ifferenciate:
1) ~f psum=psoll
Each signal receives the requested number of bits:
Z (k) _ p (k)
,., ::.::., . , ;.....:.. .,, ::.'.': , .. . -.:: _ ~.~: . ..!::~:. >.:.,
,.:,:, , ~, ~ .'":: . , .:;~.~::.:'. .. ~:-:,.:; , , ,,:.;. . '~:~':' ' :.;.
~:~.::' , ,.:, , ... ~ . :~: ,.::..
:' ..:.;: '.,..,:., H , .::: , .."..... _ .:..-:.. ~. '.,: . ':... ,
,:,...,..
.. ;..~., . . ;.::.. ~...n ,;..,t., . ~:......_. -. , ...
W ::.;.-n.~
..:.',.:. ,..::.: .., ~:.:' ,..:"'.,' :::'' , , . .'.. .. . . ~.,:~:~..:
.1. . v:.~. :. '::~: . :::
.
' . ,: . ~..' ..,: . '.;'.:'.":.....
.:. '....'..... "'. ...,:". .. .:'' ".,'v .. .:~
G o.. ~~.,..~i"~.~::,:~ ..::.~ _:..
7
1
1.. .: ~
! " 4 , .
y
n1.
,.'.G' .,t.r
,J:
' J
/
. .7 ...,.
. .,C.. . .
:.,', :u .. , :;:,. :.: ., .. '..,: ...., . .. . . .;..,.. v ..
',, , ..... :.~: ~ . . : : ,. , t ,:-'u -i-
. . ";. ' . ~ ..; S :. .v'v : ..'. ., ..
I ~ ': J. ~;':. .. ., .. ,
' ' . .' ~ . :~.'. : ,... .' ~ ~': , . : ~ ~ ! .. ": .. w" : . , . '..: .
' ~:...;' . ,.;...', '...:, . . .:: ; '.. . '..:., . ,. .
. ' ' , :: ..... .:, ..,. . ~:. .
::' ::' ' . : t . ,:: , , , . ..;.:. . F . ..:
' ':
'
. .
:.J:: ,.. .....:... . .
...,. .;. ' . , ':': ,:~: ',','.. .. ..:,._ '. .. :.
';.:-: 0 .':. ::. ..,.,..,';. . '..::.v>..'. :?: :.~
',',. ' .~::. ,. :...,..,,....,..,:.,,':..:.~:. ... ;:
, ::...., ' ' :~:;, ...n .~.' . ..".. . ~, f Y .. !~~'.. ",.:: .....
.. ....,., .::..':. .. ' .~,~.:'; .. ,:.J: ... :,..:.. ,:,. ':..._
: .;~: ..'.~'.. .:::~ : :'.S' . y:.~.....:.. .',~"' J, :J. ;.::~ .
. .':::':... :.. ~..,-: ,'
',"'..:;,: .;:.~.' ..::':? ,,. .;....... . v,:
~,y'. ...,, .".: .:.:... .','~. . ....
',' ., ~~:: W y ~~~'-. ..,. .'. . . .. '.~ ::'. .:; :::: .'..; : ;.
,..,. ~~ J ~.:.; . :'':~. ...:.'.:. '. .
',.., y :~: ' - .,.: . ,. . . ..
' ::.. '
'. :
~
g,
,. . . .
, ,.,. .. ,., .. .. . , .
:
'..:. , :: w.~ ' . '.. , .
.,...:
~ :''.:. ...; ..v.:~ ':'. :..
". , , , . . .
i ., ......... ,...:.'.; .
. . . .. :.: ... : ... . ..
.'. Y :na :.' . .
,.,,"... ... :.::..:: .. .
',:. ' .:~. . .".,:.: :~'..
- J-. ':'..:.:.,.Y.':.-:.~,:
. :'..:;,.. ". ..;..:" '-m:'
. .:~ .w:: "_..,., ...
,. .,. ,.. , ,... . ..,:.;.
,.....~.~.., , . :,i: J ~.,...
. . : ~.:: . .. :. . :..,:
..: .,... :.':~~. :' .
-10-
~~;~c,~rlr~
~ im (~
2) Tf psum<psoll
Each signal receives more than the requested number of
bits:
z(k) - (psoll/psum) * p(k)
e.g., K=2, psoll=1600, p(1)=540,p(2)=660
psum=1200
z(1) - 1600/1200 * 540 = '720 (180 bits more)
z(2) - 1600/1200 * 660 = 880 (220 bits more)
3) If psoll>psum
Each signal receives less than the requested number of
bits:
a) for OCF:
z(k) - (psoll/psum) * p(k)
b) for PCM:
The minimum number of bits for each signal must not be
undercut:
z(k) - p0 + ( (psoll-K*p0) ) * (p(k)-p0)
e.g., K=2, psoll=1600, p0=500, p(1)=600,
p(2)=1200
then Psum=1800
z(1)=500+(1600-2*500)/(1800-2*500)*(600--500)=575
(25 bits less) '
z(2)=500+(1600-2*500)/(1800-2*500)*(1200-500)=1025
(175 bits less)
In order to correct the permissible interference, the
following differenciation is necessary if p bits are
requested for each signal but z bits are alloted:
1) If the alloted number of bits equals the requested
number: no correction is required.
2) If more bits were alloted than were requested:
11
For OCF:
no correction is required.
For PCM:
The number of bits at disposal for quantization in each
section is increased by (z-p)/512.
3) 2f the number of bits being alloted is less than being
requested:
For OCF:
ss(i,b) - s(i*b) + (z-p0)/(p-o0 * (ss(i,b)-s(i,b))
for p>p0
ss (i,b) - s (i,b) for p<=p0
For PCM:
The number of bits at disposal for quantization in each
section are increased by (z-p)/512.
Tn the case of PCM, a rounding off bit per ATW to an
integer is required: in order to do this all bits/ATW
are rounded off to the next lowest integer and the
resulting bit sum is determined therefrom.
rounded off: 4 5 3 2
*Width 16 30 2~ 2~
still to be alloted: 10 bits
+1 +1
Result: 5 6 3 2
The present invention has been described hereinbefore
using preferred embodiments, There are, of course, very
many different variations possible within the scope of the
overall inventive idea:
A fixed overall block length can be employed, with filling
bits being used or there is a transfer of not yet ended
coders. Furthermore, a flexible block length can be
employed which prescribes a maximum block length and in
addition time averaging occurs.
