EP0610282B1 - Process for simultaneously transmitting audio signals from n-signal sources - Google Patents

Abstract

A process is disclosed for simultaneously transmitting signals from N-signal sources over a corresponding number of transmission channels. The individual signals are subdivided into blocks and the blocks are converted into spectral coefficients by transformation or filtering, and the latter are then subjected to a data reduction process. The invention is characterized in that the blocks belonging to the individual signals are subdivided into section. The momentary sections of all signals are processed together, the admissible disturbance is determined for each section by using a perception-specific model and the momentarily required total transmission capacity is calculated. The allocation of the maximum available transmission capacity for each individual signal is calculated from the total available transmission capacity and the total momentarily required transmission capacity. Each individual signal is coded and transmitted with the thus determined capacity.

Description

Technical field

The invention relates to a method for simultaneous Transmission of audio signals from N signal sources via a corresponding number of transmission channels in which the individual Signals divided into blocks and the blocks by one Transformation or filtering implemented in spectral coefficients which are subjected to a data reduction process, and the total available transmission capacity and the currently required total transmission capacity the allocation to maximum available standing transmission capacity for each individual signal calculated and each individual signal determined with this Capacity is encoded and transmitted.

State of the art

Procedures in which the individual (time) signals in Blocks divided and the blocks through a transformation or filtering implemented in spectral coefficients which in turn is a data reduction process subjected or coded accordingly for data reduction are known. For example, on the review articles "Perceptual Audio-Coding" by Jörg Houpert in Studio Technique or the article "Data Diet, Data reduction with digitized audio signals " directed by Stefanie Renner in Elrad, 1991. On these summary articles as well as the PCT published application Incidentally, WO 88/01811 is used to explain all here Terms and process steps not described in detail expressly referred.

In a number of cases, it is now necessary Signals from several signal sources simultaneously a corresponding number of transmission channels transfer. The simplest example of this is the Transmission of stereo signals over two transmission channels called.

When transmitting signals from N signal sources over a corresponding number of transmission channels now the problem of dimensioning the Transmission channels:

If you dimension each individual transmission channel so that he transmit the "maximum bit stream" can, so "on average" remains comparatively much transmission capacity unused.

Now it’s known from digital phone technology, when transmitting signals from a variety of Signal sources over a corresponding number of transmission channels the transmission channels for only one to interpret "medium demand" and on individual channels short-term increased demand through assignment from others Equalize channels. The assignment is made exclusively via the signal statistics.

The following references refer to the state of the art "a digital speech interpolation method with predictive control Word division "by Dr. H. Gerhäuser (1980), "a digital speech interpolation method with current priority assignment ", by R. Woitowitc (1977) or "a digital speech interpolation method with block priority allocation " by G.G. Klahnenbucher (1978) referenced.

It has been recognized that in digital telephone technology Common procedures to compensate for fluctuating needs when transmitting a variety of audio signals over a corresponding number of transmission channels then no good ones Deliver results when the digital to be transferred Signals before data reduction, for example have been subjected to the so-called OCF process.

From the contribution by Kuei Yung Kou et al, "Digital Speech Interpolation for Variable Rate Coders with Application to Subband Coding ", IEEE Transaction on Communications, Vol. COmm 33, No. 10, Oct. 1985, pp. 1100-1108, a process for simultaneous transmission of signals from n signal sources forth. Not about a given transmission quality of the Siganel to fall below, iterative measurements and fitting procedures applied, with the associated after each individual bit allocation S / N ratio is measured. Corresponds to the measured S / N value a predetermined threshold value, then a corresponding Signal adaptation carried out. However, it turned out that by the mere detection of the S / N ratio a meaningful Human quality assessment Hearing audio signals is not possible.

Presentation of the invention

The invention has for its object a method for the simultaneous transmission of audio signals from N signal sources over a corresponding number of transmission channels specify with which "data-reduced Signals "via transmission channels that are only for an "average need" are dimensioned without perceptible, i.e. for example, audible losses Signal capacity can be transmitted.

An inventive solution to this problem is in Claim 1 specified. Developments of the invention are the subject of the subclaims.

The invention is based on the basic idea of Compensate for fluctuating needs at the same time Transmission of audio signals from N signal sources via a corresponding number of transmission channels Allocation to the individual signals not according to statistical Aspects, but already in the process step in which the signals for data reduction are encoded, the fluctuating demand to compensate for corresponding measures.

Fig. 1

ein Blockdiagramm zur Erläuterung des erfindungsgemäßen Verfahrens, und

Fig. 2a und 2b

den erfindungsgemäßen Signalaufbau.

This basic idea according to the invention is explained in more detail below with reference to an exemplary embodiment under the drawing, in which:

Fig. 1

a block diagram for explaining the method according to the invention, and

2a and 2b

the signal structure according to the invention.

In the method according to the invention, the individual Audio signals divided into blocks and the blocks through a transformation or filtering in spectral coefficients implemented. To compensate for fluctuating needs become those belonging to the individual signals Blocks divided into sections, and the current one Sections of all signals processed together. This is shown in FIG. 1 by corresponding "function blocks" represented graphically.

Using a perception-specific model, for example when transferring Audio signals can be a psycho-acoustic model the permitted fault is determined for each section and from this the requirement for currently required Total transmission capacity calculated. The calculation the total transmission capacity, i.e. the necessary Number of bits, occurs for all blocks at the same time. From the total available transmission capacity and the currently required total transmission capacity the allocation to the maximum available transmission capacity calculated for each individual signal. With the "number of bits" allocated for each individual signal the individual signal is encoded and accordingly the transmission of this single signal. Here in the simplest case, each is balanced required transmission capacity only between the Channels.

In the training specified in claim 2 a reserve of transmission capacity, a so-called Bit reservoir available from which in the event that the total transmission capacity required on average available transmission capacity exceeds an allocation of transmission capacity is made.

This bit reservoir is always filled up when the requested transmission capacity is less than that available transmission capacity is (Claim 3).

In any case, it is necessary that - to get one to prevent large growth of the bit reservoir - in in the event that the transmission capacity is very much smaller than the available transmission capacity is a forced allocation of bits to the individual channels takes place (claim 4). This compulsory allocation is preferably done only to the Channels or signal sources that registered a need that are greater than a medium need. A much greater need than the average Namely, need means that these signals are essential are more difficult to code than normal signals.

In any case, it is preferred according to claim 9 if from all separately coded signals from the signal sources an entire block is formed that is on a solid Area that contains information from which the Separation of the signals can be determined, as well consists of several areas of flexible length that the encoded signals. This is shown schematically in Fig. 2a shown.

Another saving in transmission capacity is obtained one recognizes that the same input signals and through a suitable transmission format only once are transferred (claim 6). This is shown schematically in Fig. 2b shown.

In any case, it is possible to find the one currently required To determine transmission capacity exactly or only to estimate (claims 7 and 8).

In addition, it is possible to use the invention Execute procedures largely in parallel. This is it is preferred if, according to claim 11, the coding of the Individual signals already during the calculation of the allocation the transmission capacity for each signal.

Another preferred implementation of the invention The basic idea is stated in claim 11:

If the required transmission capacity is available standing transmission capacity exceeds and no allocation can be made from the bit reservoir, so it is possible to change the value of the allowed interference for raise all signals so that the required Total transmission capacity the available Transmission capacity does not exceed (claim 11).

The following is a numerical example of a Procedure for audio signals can be specified. It is expressly pointed out that the invention Basic idea is not limited to audio signals is; rather you can also use video signals or another, a perception-specific assessment underlying signals are treated similarly.

Example of a possible procedure for audio signals:

1) Das Audio-Signal y wird in an sich bekannter Weise in Abtastwerte (y(t) zerlegt, die digitalisiert werden. Die digitalisierten Abtastwerte werden in Blöcke der Länge 2n zerlegt, die bei dem gewählten Ausführungsbeispiel überlappende Blöcke mit der Überlappung n sind: x(k,b) = y(b*n+k)    für k=O..2n (b Blocknummer).

2) Jeder Block der Länge n wird durch eine Transformation, beispielsweise eine Fast-Fourier-Transformation oder eine Cosinus-Transformation in Spektralkoeffizienten transformiert: x(j,b) = SUM(1=0..2n; x(1,b).f(1).cos(pi.(21+1+n) (2j+1) / (4n)) ) für j=O..n mit f(1) = sqrt(2).sin (pi - (1+O.5) / (2n) )

3) Jeder der Blöcke wird in Abschnitt zerlegt und die Energiedichte für jeden Abschnitt berechnet: E(i,b) = ( SUM(k=a(i)+1..a(i+1 X(k,b)^2 ) ) / (a(i+1)-a(i)) für i=1 ... c, wobei die Koeffizienten a(i) aus der untenstehenden Tabelle 1 entnommen werden.

4) Für jeden.Abschnitt wird mit einem geeigneten psycho-akustischen Modell, bezüglich dem auf die Literatur verwiesen wird, die erlaubte Störung berechnet. Aus der erlaubten Störung ergibt sich die Maskierung zwischen den Bändern T(i,b) = MAX(k=1 ... i-1; E(k,b) z(i-k) ) die Maskierung im Band: s(i,b) = max ( E(i,b) e(i) , T(i,b) ) und die Maskierung zwischen den Blöcken: ss(i,b) = max ( s(i,b-1(/16 , s(i,b) ) anschließend erfolgt für jeden Block die Berechnung der benötigten Bitzahl.

5) Berechnung der nötigen Bitzahl für den Block:

a) für eine Codierung wie bei OCF (Huffmancodierung): p = pO + SUM(i=1..C; (a(i+1)-a(i) . (s(i,b)/ss(i,b) ) )

b) für PCM Codierung (SNR = 6dB/bit) :
Für jeden Abschnitt wird ein Skalenfaktor und die Anzahl der Bit pro Abtastwert als zusätzliche Information übertragen p = pO + SUM(i=1..c; (a(i+1)) . 10/6 . log( E(i,b) / ss(i,b) ) )

Let y (t) be samples of the audio signal.

1) The audio signal y is broken down in a manner known per se into samples (y (t) which are digitized. The digitized samples are broken down into blocks of length 2n, which in the selected exemplary embodiment are overlapping blocks with the overlap 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 a transformation, for example a Fast Fourier transformation or a cosine transformation: x (j, b) = SUM (1 = 0..2n; x (1, b) .f (1) .cos (pi. (21 + 1 + n) (2j + 1) / (4n))) for j = O..n with f (1) = sqrt (2) .sin (pi - (1 + O.5) / (2n))

3) Each of the blocks is broken down into sections and the energy density calculated for each section: E (i, b) = (SUM (k = a (i) + 1..a (i + 1 X (k, b) ^ 2)) / (a (i + 1) -a (i)) for i = 1 ... c, the coefficients a (i) being taken from Table 1 below.

4) For each section, the permitted disorder is calculated using a suitable psycho-acoustic model, with reference to which reference is made to the literature. The masking between the bands results from the permitted interference T (i, b) = MAX (k = 1 ... i-1; E (k, b) z (ik)) the masking in the band: s (i, b) = max (E (i, b) e (i), T (i, b)) and the masking between the blocks: ss (i, b) = max (s (i, b-1 (/ 16, s (i, b)) the required number of bits is then calculated for each block.

5) Calculation of the necessary number of bits for the block:

a) for coding as with OCF (Huffman coding): p = pO + SUM (i = 1..C; (a (i + 1) -a (i). (s (i, b) / ss (i, b)))

b) for PCM coding (SNR = 6dB / bit):
For each section, a scale factor and the number of bits per sample are transmitted as additional information p = pO + SUM (i = 1..c; (a (i + 1)). 10/6. log (E (i, b) / ss (i, b)))

n =

512

c =

23

p0 =

1200 für OCF (mittlere Bitzahl pro Block)

pO =

345 für PCM (Skalenfaktoren: 10 Bit/Abschnitt, Codierung der Anzahl der Quantisierungsstufen: 5 Bit/Abschnitt

In the following, the meaningful values for the individual quantities or constants are to be given in the form of tables:

n =

512

c =

23

p0 =

1200 for OCF (average number of bits per block)

pO =

345 for PCM (scale factors: 10 bits / section, coding of the number of quantization levels: 5 bits / section

The bit numbers are then assigned to the individual signals. For this purpose, it is assumed that the encoding of the K input signals k (k) bits are required as the number of bits available to p _should be. psum = SUM (p) k))

1) wenn psum=psoll
Jedes Signal bekommt die angeforderte Bitzahl:
z(k) = P(k)

2) Wenn psum<psoll
Jedes Signal bekommt mehr als die angeforderte Bitzahl: z(k) = (psoll/psum) . p(k) z.B. K=2, psoll=1600, p(1)=540, p(2)=660 psum=1200 z(1) = 1600/1200 . 540 = 720 (180 bit mehr) z(2) = 1600/1200 . 660 = 880 (220 bit mehr)

3) Wenn psoll > psum
Jedes Signal bekommt weniger als die angeforderte Bitzahl:

a) für OCF: z(k) = (psoll/psum) . p(k)

b) für PCM:
Die Mindestbitzahl für jedes Signal darf dabei nicht unterschritten werden: z(k) = pO + ( (psoll-K pO) ) . (p(k)-pO)

z.B. K=2, psoll=1600, p0=500, p(1)=600,
p(2)=1200
dann ist Psum=1800 z(1)=500+(1600-2.500)/(1800-2.500) (600-500)=575 (25 bit weniger) z(2)=500+(1600-2.500(/(1800-2.500) (1200-500)=1025 (175 bit weniger) Now a case distinction is necessary:

1) if psum = psoll
Every signal gets the requested number of bits:
z (k) = P (k)

2) If psum <psoll
Each signal gets 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 bit more) z (2) = 1600/1200. 660 = 880 (220 bit more)

3) If psoll> psum
Each signal gets 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) = pO + ((psoll-K pO)). (p (k) -pO)

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 bit less) z (2) = 500 + (1600-2.500 (/ (1800-2.500) (1200-500) = 1025 (175 bit less)

1) Wenn zugeteilte Bitzahl gleich der angeforderten: keine Korrektur nötig.

2) Wenn mehr Bits zugeteilt wurden als angefordert:
Für OCF:
keine Korrektur nötig.
Für PCM:
Die Anzahl der für die Quantisierung in jedem Abschnitt zur Verfügung stehenden Bit wird um /z-p)/512 vermehrt.

3) Wenn weniger Bits zugeteilt wurden als angefordert: Für OCF: ss(i,b) = s(i,b) + (z-p0)/(p-oO . (ss(i,b)-s(i,b)) für p>p0 ss(i,b) = s(i,b)    für p<=p0

Für PCM:
Die Anzahl der für die Quantisierung in jedem Abschnitt zur Verfügung stehenden Bit wird um (z-p) / 512 vermehrt.The following case distinction is necessary to correct the permitted interference if p-bits are requested for each signal, but z-bits are allocated:

1) If the number of bits allocated is equal to the requested number: no correction necessary.

2) If more bits have been allocated than requested:
For OCF:
no correction necessary.
For PCM:
The number of bits available for quantization in each section is increased by / zp) / 512.

3) If fewer bits were allocated than requested: For OCF: ss (i, b) = s (i, b) + (z-p0) / (p-oO. (ss (i, b) -s (i, b)) for p> p0 ss (i, b) = s (i, b) for p <= p0

For PCM:
The number of bits available for quantization in each section is increased by (zp) / 512.

With PCM, one rounding bit is per whole ATW Number required: For this, all bits / ATW are first opened rounded off the next lowest whole number and the one from it resulting bit total is determined.

If bits are still available, the first Pass each from the lowest bands Tape one bit / ATW provided until the available number of bits is reached.

Example:

104 bits are available Section: 1 2nd 3rd 4th Width: 4th 6 8th 12th Bit / ATW: 4.2 5.2 3.4 2.4 undounded: 4th 5 3rd 2nd *Width 16 30th 24th 24th still to be distributed: 10 bits +1 +1 Result: 5 6 3rd 2nd

The invention is based on exemplary embodiments have been described. Within the general The idea of the invention is of course the most varied Variations possible:

So it is possible to have a fixed total block length use fill bits or a Transfer to coder that has not yet ended. Further it is possible to use a flexible block length at which a maximum block length is specified and there is also a time averaging.

Claims (11)

1. A process for simultaneous transmitting of audio signals of N signal sources via a corresponding number of transmission channels,

   in which the individual signals are divided into blocks and said blocks are transformed into spectral coefficients by transformation or filtering, said spectral coefficients undergoing a data reduction process, and in which the allotment of the maximum disposable trans- mitting capacity is calculated for each individaul signal from the entire disposable transmitting capacity and the currently required overall transmitting capacity and each individual signal whose capacity was determined in this manner is encoded and transmitted,

   characterized by the following features:

   said blocks belonging to said individual signals are divided into sections,

   the respective current sections of all signals are processed simultaneously,

   the permissible interference for each section is determined utilizing a perception-specific model and a request of currently required overall transmitting capacity is calculated.
2. Process according to claim 1,
   characterized by there being a reserve of transmitting capacity (bit reservoir) from which an allotment occurs if the required overall transmitting capacity exceeds the average transmitting capacity at disposal.
3. Process according to claim 2,
   characterized by said bit reservoir being filled if the requested transmitting capacity is smaller than the transmitting capacity at disposal.
4. A process according to claim 3,
   characterized by a forced allotment occurring in order to prevent said bit reservoir from increasing too greatly if the requested transmitting capacity is very much smaller than the transmitting capacity at disposal.
5. A process according to claim 4,
   characterized by said forced allotment only occurring if there is a need greater than the average need.
6. A process according to one of the claims 1 to 5, characterized by identical input signals being recognized and being transmitted only once by a suited transmitting format.
7. A process according to one of the claims 1 to 6, characterized by the determination of the currently required transmitting capacity occuring accurately.
8. A process according to one of the claims 1 to 6, characterized by said determination of said currently required transmitting capacity being only estimated.
9. A process according to one of the claims 1 to 8, characterized by an overall block being formed from all the separately encoded signals from the signal sources, said overall block being composed of a fixed section containing information from which the separation of said individual signals can be determined and composed of several regions of flexible length.
10. A process according to one of the claims 1 to 9, characterized by said encoding of said individual signals already occurring during the calculation of the allotment of the transmitting capacity for each signal.
11. A process according to one of the claims 1 to 10, characterized by, if the requested number of bits exceeds the overall number of bits at disposal, the permissable interference for all the signal sources being increased so that a reduced bit request is yielded.

Images