RU2777660C2 - Method and device for formation from representation of hoa signals in domain of mixed representation coefficients of mentioned hoa signals in spatial domain/coefficient domain - Google Patents

Abstract

FIELD: data processing.

SUBSTANCE: signal vector in a coefficient domain is divided into a signal vector in the coefficient domain, having a constant number of higher order ambiophonic sound (hereinafter – HOA) coefficients, and a signal vector in the coefficient domain, having an alternating number of HOA coefficients. The vector having the constant number of HOA coefficients is converted into a corresponding signal vector in a spatial domain. The signal vector having the alternating number of HOA coefficients in the coefficient domain is adaptively normalized and multiplexed with the signal vector in the spatial domain.

EFFECT: provision of a possibility of the formation from a representation in a domain of mixed representation coefficients in a spatial domain/coefficient domain, in which a number of HOA signals can be variable.

3 cl, 8 dwg

Description

The technical field to which the invention belongs

The present invention relates to a method and apparatus for generating from a coefficient domain representation of HOA signals a mixed spatial/coefficient domain representation of said HOA signals, in which the number of HOA signals can be variable.

State of the art

Referred to as HOA (Higher Order Ambisonics), High Order Ambiophonic Sounds are a mathematical description of a two- or three-dimensional sound field. The sound field can be captured by a set of microphones, generated from synthetic sound sources, or a combination of both. HOA can be used as a transport format for 2D or 3D surround sound. Unlike loudspeaker-based surround sound presentations, the advantage of HOA is the reproduction of the sound field in different speaker setups. Therefore, HOA is suitable for a universal audio format.

The spatial resolution of a HOA is determined by the order of the HOA. This order specifies the number of HOA signals that describe the sound field. There are two representations for HOA, which are called the spatial domain and coefficient domain, respectively. In most cases, HOAs are initially represented in coefficient domain and such a representation can be converted to spatial domain by matrix multiplication (or transformation) as described in EP 2469742 A2. The spatial domain consists of the same number of signals as the coefficient domain. However, in the spatial domain, each signal is relative to a direction, with the directions evenly distributed over the unit sphere. This facilitates the analysis of the spatial distribution of the HOA representation. Representations in the coefficient domain, just like representations in the spatial domain, are representations in the time domain.

Disclosure of invention

In the following, in general, the goal is to use as much as possible for the PCM transmission of HOA representations of the spatial domain to provide the same dynamic range for each direction. This means that discrete PCM samples of HOA signals in the spatial domain must be normalized to a range of predefined values. However, the disadvantage of such normalization is that the dynamic range of HOA signals in the spatial domain is less than in the coefficient domain. This is due to the transform matrix, which generates the signal in the spatial domain from the signals in the coefficient domain.

In some applications, HOA signals are transmitted in the gain domain, such as in the processing described in EP 13305558.2, in which all signals are transmitted in the gain domain because a constant number of HOA signals and a variable number of additional HOA signals must be transmitted. However, as already mentioned above and shown in EP 2469742 A2, transmission in the coefficient domain is not advantageous. As a solution, a constant number of HOA signals may be transmitted in the spatial domain, and only additional HOA signals with a variable number are transmitted in the coefficient domain. The transmission of additional HOA signals in the spatial domain is not possible, since the time varying number of HOA signals will result in time varying transformation matrices from the coefficient domain to the spatial domain, and discontinuities may occur in all signals in the spatial domain, which are suboptimal for subsequent perceptual encoding of PCM signals.

To ensure that these additional HOA signals are transmitted without exceeding the range of predetermined values, reversible normalization processing can be used, which is designed to prevent such signal gaps, and which also achieves efficient transmission of inversion parameters.

Regarding the dynamic range of the two HOA representations and normalization of HOA signals for PCM encoding, the following concludes whether such normalization should take place in the coefficient domain or in the spatial domain.

In the coefficient time domain, the HOA representation consists of successive frames consisting of N signal coefficients d _n ( k ), n =0, ..., N -1 , where k denotes the sample index and n denotes the signal index.

These coefficient signals are assembled into a vector d ( k )=[ d ₀ ( k ), …, d _{N -1} ( k )] ^T to obtain a compact representation.

The transformation to the spatial domain is performed by means of an N x N transformation matrix

as defined in EP 12306569.0, see the definition for Ξ _GRID in conjunction with equations (21) and (22).

The vector w ( k )=[ w ₀ ( k ), …, w _{N -1} ( k )] ^T in the spatial domain is obtained from w ( k ) = Ψ ^-1 d ( k ) , (1)

where Ψ ^-1 is the inverse of the matrix Ψ .

The reverse transformation from the spatial domain to the coefficient domain is performed by

d ( k )=Ψ w ( k ). (2)

If the range of values of discrete samples is set in one area, then the transformation matrix Ψ automatically sets the range of values of another area. The term ( k ) for the kth discrete sample is omitted further.

Because the HOA representation is actually rendered in the spatial domain, the value range, loudness, and dynamic range are set in that domain. The dynamic range is given by the bit resolution of the PCM encoding. In this application, "PCM encoding" means converting discrete samples in floating point representation to discrete samples in integer representation in fixed point notation.

For PCM encoding of a HOA representation, the N signals in the spatial domain must be normalized to the range of -1 ≤ w _n < 1 so that they can be scaled to the maximum PCM value W _max and rounded to a fixed-point integer PCM notation

(3)

(3)

Note: This is a generalized representation of PCM encoding.

The range of values for discrete samples in the coefficient domain can be calculated using the infinite matrix norm of the matrix Ψ , which is defined as follows

, (4)

, (four)

and the maximum absolute value in the spatial domain w _max = 1 for - ║ Ψ ║ _∞ w _max ≤ d _n < ║ Ψ ║ _∞ w _max . Since the value of ║ Ψ ║ _∞ is greater than '1' for the definition of the matrix Ψ used, then the range of values of d _n increases.

The reverse move means that normalization to ║ Ψ ║ _∞ is necessary for PCM-coding of signals in the coefficient domain since -1 ≤ d _n / ║ Ψ ║ _∞ < 1 . However, this normalization reduces the dynamic range of the signals in the gain domain, which will result in a reduced quantization signal-to-noise ratio. Therefore, PCM coding of signals in the spatial domain should be preferred.

The problem to be solved by the present invention is how to convey in the coefficient domain a portion of the desired HOA signals in the spatial domain using normalization without reducing the dynamic range in the coefficient domain. Additionally, the normalized signals should not contain signal level jumps so that they can be perceptually encoded without the loss of quality caused by the jumps. This problem is solved using the methods disclosed in paragraphs 1 and 6 of the claims. Devices that use these methods are disclosed in paragraphs 2 and 7 of the claims, respectively.

In principle, the inventive generation method is suitable for generating from the representation of HOA signals in the coefficient domain a mixed representation of the mentioned HOA signals in the spatial domain/coefficient domain, and the number of said HOA signals can be variable in time in successive frames of the coefficients, differing in steps, which:

dividing the coefficient domain HOA signal vector into a first coefficient domain signal vector having a constant number of HOA coefficients and a second coefficient domain signal vector having a time-varying number of HOA coefficients;

converting said first coefficient domain signal vector to a corresponding spatial domain signal vector by multiplying said coefficient domain signal vector by an inverse transform matrix;

PCM-encoding said vector of signals in the spatial domain to obtain a vector of PCM-encoded signals in the spatial domain;

- normalizing said second signal vector in the coefficient domain by a normalization factor, wherein said normalization is an adaptive normalization with respect to the range of current values of the HOA coefficients of said second signal vector in the coefficient domain, and with said normalization the range of available values for the HOA coefficients of the vector is not exceeded , and while under said normalization, a uniformly continuous transfer function is applied to the coefficients of the current second vector in order to continuously change the gain within this vector from the gain in the previous second vector to the gain in the next second vector, and moreover, said normalization provides auxiliary information for the corresponding denormalization on the decoder side;

- PCM-encoding said vector of normalized signals in the coefficient domain to obtain a vector of PCM-encoded and normalized signals in the coefficient domain;

- multiplexing said vector of PCM-encoded signals in the spatial domain and said vector of PCM-encoded and normalized signals in the coefficient domain.

In principle, the inventive generating device is suitable for generating from the coefficient domain representation of HOA signals a mixed spatial/coefficient domain representation of said HOA signals, wherein the number of said HOA signals can be time-varying in successive coefficient frames, said device comprising into yourself:

means configured to divide the coefficient domain HOA signal vector into a first coefficient domain signal vector having a constant number of HOA coefficients and a second coefficient domain signal vector having a time-varying number of HOA coefficients;

means configured to convert said first coefficient domain signal vector to a corresponding spatial domain signal vector by multiplying said coefficient domain signal vector by an inverse transformation matrix;

means configured to PCM encode said vector of signals in the spatial domain to obtain a vector of PCM encoded signals in the spatial domain;

- means configured to normalize said second signal vector in the coefficient domain by a normalization factor, wherein said normalization is an adaptive normalization with respect to a range of current values of HOA coefficients of said second signal vector in the coefficient domain, and with said normalization, the range of available values for HOA -coefficients of the vector is not exceeded, and moreover, with the mentioned normalization, a uniformly continuous transfer function is applied to the coefficients of the current second vector in order to continuously change the gain inside this vector from the gain in the previous second vector to the gain in the next second vector, and at the same time, the mentioned normalization provides ancillary information for appropriate denormalization at the decoder side;

means configured to PCM-code said vector of normalized signals in the coefficient domain to obtain a vector of PCM-encoded and normalized signals in the coefficient domain;

means for multiplexing said vector of PCM-encoded signals in the spatial domain and said vector of PCM-encoded and normalized signals in the coefficient domain.

In principle, the inventive decoding method is suitable for decoding a mixed representation of encoded HOA signals in the spatial/coefficient domain, in which the number of said HOA signals can be time-varying in successive frames of coefficients, and in which said mixed representation of the encoded HOA signals in the spatial area/region of coefficients is formed according to the above-mentioned inventive method of formation, while said decoding includes the steps of:

demultiplexing said multiplexed vectors of PCM-encoded signals in the spatial domain and PCM-encoded and normalized signals in the coefficient domain;

converting said vector of PCM-encoded signals in the spatial domain to a corresponding vector of signals in the coefficient domain by multiplying said vector of PCM-encoded signals in the spatial domain by said transformation matrix;

- denormalizing said vector of PCM-encoded and normalized signals in the coefficient domain, wherein said denormalizing step includes the steps of:

-- calculate, using the appropriate exponent (exponent) e _n ( j -1) from the received auxiliary information and the recursively calculated value g _n ( j -2) of the gain, the vector h _n ( j -1) of the transfer function, and save the gain value g _n ( j -1) of the gain for the corresponding processing of the next vector of PCM-encoded and normalized signals in the area of the coefficients to be processed, while j is a variable (running, running) index of the input matrix of HOA signal vectors;

-- apply the corresponding reciprocal gain value to the current PCM-encoded and normalized signal vector to obtain the corresponding PCM-encoded and denormalized signal vector;

combining said coefficient-domain signal vector and the denormalized coefficient-domain signal vector to obtain a combined coefficient-domain HOA signal vector, which may have a variable number of HOA coefficients.

In principle, the inventive decoding apparatus is suitable for decoding a mixed representation of encoded HOA signals in the spatial/coefficient domain, in which the number of said HOA signals can be time-varying in successive frames of coefficients, and in which said mixed representation of the encoded HOA signals in the spatial domain /coefficient region is formed according to the above-mentioned generation method, wherein said decoding device includes:

means for demultiplexing said multiplexed vectors of PCM-encoded signals in the spatial domain and PCM-encoded and normalized signals in the coefficient domain;

means for converting said vector of PCM-encoded spatial domain signals to a corresponding vector of coefficient-domain signals by multiplying said vector of PCM-encoded spatial domain signals by said transformation matrix;

- means configured to denormalize said vector of PCM-encoded and normalized signals in the coefficient domain, said denormalization comprising:

-- calculation, using the appropriate exponent e _n ( j -1) of the received auxiliary information and the recursively calculated value g _n ( j -2) of the gain, the vector h _n ( j -1) of the transfer function, and the ability to save the value g _n ( j -1) gain for the corresponding processing of the next vector of PCM-encoded and normalized signals in the area of the coefficients to be processed, while j is a variable index of the input matrix of vectors of HOA signals;

-- applying the corresponding reciprocal gain value to the current PCM-encoded and normalized signal vector to obtain the corresponding PCM-encoded and denormalized signal vector;

means configured to combine said coefficient domain signal vector and the coefficient domain denormalized signal vector to obtain a combined coefficient domain HOA signal vector, which may have a variable number of HOA coefficients.

Preferred additional embodiments of the present invention are disclosed in the respective dependent claims.

Brief description of the drawings

Exemplary embodiments of the present invention are described with reference to the accompanying drawings, in which:

1 shows a PCM transmission in the spatial domain of the original HOA representation in the coefficient domain;

2 shows the combined transmission of the HOA representation in the coefficient domain and the spatial domain;

FIG. 3 shows the combined transmission of a coefficient-domain and spatial-domain HOA representation using block adaptive normalization for coefficient-domain signals;

4 shows adaptive normalization processing for a HOA signal x _n ( j ) represented in the coefficient domain;

5 shows the transfer function used to smoothly transition between two different gain values;

Fig. 6 shows adaptive denormalization processing;

Figure 7 shows the frequency FFT spectrum of the transfer functions h _n ( l ) using various exponents e _n , in which the maximum amplitude of each function is normalized to 0 dB;

8 shows exemplary transfer functions for three successive signal vectors.

Description of Embodiments

Regarding the PCM encoding of the HOA representation in the spatial domain, it is assumed that (in floating point representation) -1 ≤ w _n < 1 is performed so that the PCM transmission of the HOA representation can be performed as shown in FIG. . In the transform step or step 11 at the input of the HOA encoder, the signal d in the coefficient domain of the current input frame of the signal is converted to the signal w in the spatial domain using equation (1). In the PCM encoding step or step 12, the floating-point discrete samples w are converted to PCM-encoded integer discrete samples w' in fixed-point notation using equation (3). In step or step 13 of the multiplexer, the discrete samples w' are multiplexed into the HOA transmission format.

The HOA decoder demultiplexes the signals w' from the received HOA transmission format in demultiplexer step or step 14, and reconverts them in step or step 15 to coefficient domain signals d ' using equation (2). This inverse transform increases the dynamic range d ' such that such a spatial-to-coefficient-domain conversion always includes a format conversion from integer (PCM) to floating point.

The standard HOA transmission of FIG. 1 will not succeed if the matrix Ψ is time-varying, which is the case if the number or index of HOA signals is time-varying for successive sequences of HOA coefficients, i.e., successive signal input frames. As mentioned above, one example of such a case is the HOA compression processing described in EP 13305558.2: a constant number of HOA signals are transmitted continuously, and a variable number of HOA signals with varying signal indices n are transmitted in parallel. All signals are transmitted in the gain region, which is not optimal, as explained above.

According to the present invention, the processing described in conjunction with FIG. 1 is extended as shown in FIG.

In step or step 20, the HOA encoder divides the HOA vector d into two vectors d ₁ and d ₂ , where the number M of HOA coefficients for vector d ₁ is constant and vector d ₂ contains a variable number K of HOA coefficients. Because signal indices n do not change over time for vector d ₁ , PCM encoding is performed in the spatial domain in steps or steps 21, 22, 23, 24 and 25 with signals corresponding to w ₁ and w' ₁ depicted in the lower signal path of FIG. .2 corresponding to steps/steps 11-15 of FIG. However, multiplexer stage/stage 23 receives an additional input d ″ ₂ , and demultiplexer stage/stage 24 in the HOA decoder provides another output d ″ ₂ .

The number of HOA coefficients, or size, K of vector d ₂ is time-varying, and transmitted HOA-signal indices n may be time-varying. This prevents transmission in the spatial domain, as a time-varying transform matrix would be required, resulting in signal gaps in all perceptually encoded HOA signals (perceptual encoding step or step not shown). However, such signal breaks should be avoided because they reduce the quality of the perceptual encoding of the transmitted signals. Thus, d ₂ must be transmitted in the coefficient domain. Due to the larger range of signal values in the coefficient domain, the signals must be scaled in step or step 26 by a factor of 1/║ Ψ ║ _∞ before PCM encoding can be applied in step or step 27. However, the disadvantage of such scaling is that the maximum absolute value for ║ Ψ ║ _∞ is a worst-case estimate whose maximum absolute sample value will not occur very often because the range of expected values is usually smaller. As a result, the available resolution for PCM coding is not used effectively and the quantization signal-to-noise ratio is low.

The output signal d ” ₂ stage/stage 24 of the demultiplexer is scaled back in stage or stage 28 using a factor ║ Ψ ║ _∞ . The resulting signal d ''' ₂ is combined in step or step 29 with the signal d ' ₁ , resulting in a decoded HOA signal d ' in the coefficient domain.

According to the present invention, coefficient-domain PCM coding efficiency can be improved by using signal-adaptive signal conditioning. However, such normalization must be reversible and uniformly continuous from one discrete sample to another discrete sample. The required block adaptive processing is depicted in FIG. j -th input matrix D ( j )=[ d ( jL+ 0) ... d ( jL + L - one)] contains L vectors d HOA signal (index j not shown in figure 3). Matrix D split into two matrices D _one and D ₂ similar to the processing in Fig.2. Processing for D _onein steps or steps 31-35 corresponds to the spatial domain processing described in conjunction with FIG. 2 and FIG. However, coefficient-domain signal coding includes a block adaptive normalization step or step 36 that automatically adapts to a range of current signal values, followed by a PCM coding step or step 37 . Required side information to denormalize each PCM-encoded signal in the matrix D ”₂ stored and transmitted in a vector e . Vector e =[ e _{n one} … e _nK ] ^T contains one value per signal. In step or step 38 of the corresponding adaptive denormalization of the decoder on the receiving side, the signal normalization is reversed D ”₂ in D '''₂ using information from the passed vector e . Resulting signal D '''₂ combined in step or step 39 with signal D' _one, resulting in a decoded HOA signal D' in the area of coefficients.

In adaptive normalization, in step/step 36, a uniformly continuous transfer function is applied to the samples of the current input coefficient block to permanently change the gain from the last input coefficient block to the gain of the next input coefficient block. This kind of processing requires one block delay because the change in the normalization gain (normalization factor) must be detected one input coefficient block ahead. The advantage is that the amplitude modulation introduced is small, so that perceptual coding of the modulated signal has almost no effect on the denormalized signal.

As for the implementation of adaptive normalization, it is performed independently for each HOA signal D ₂ ( j ) . Signals are represented by row vectors x _n ^T matrices

where n denotes the indices of the transmitted HOA signals. x _n is transposed because it is originally a column vector, but here a row vector is required.

4 shows this adaptive normalization at step/step 36 in more detail. Processing input values are as follows:

- time-smoothed maximum value x _{n ,max,s m} ( j -2) ,

is the value g _n ( j -2) of the gain, i.e. the gain that was applied to the last gain of the corresponding block x _n ( j -2) of the signal vector,

- signal vector of the current block x _n ( j ) ,

- signal vector of the previous block x _n ( j -1) .

At the start of processing the first block x _n (0), the recursive input values are initialized to the predefined values: the coefficients of the vector x _n (-1) can be set to zero, the value g _n (-2) of the gain must be set to '1', and x _{n ,max,s m} (-2) must be set to the predefined average peak value.

After that, the value g _n ( j -1) of the gain of the last block, the corresponding value e _n ( j -1) of the vector e ( j -1) of the auxiliary information, the time-smoothed maximum value x _{n ,max,s m} ( j - 1) and the normalized vector x' _n ( j -1) of the signal are the output results of processing.

The purpose of this processing is to continuously change the gain values applied to the vector x _n ( j -one)signal, s g _n ( j -2) on the g _n ( j -one) so that the value g _n ( j -one)gain normalized vector x _n ( j )signal to a range of suitable values.

In the first processing step or step 41, each coefficient of the vector x _n ( j )=[ x _{n ,0} ( j ) ... x _{n, L -1} ( j )] of the signal is multiplied by the value g _n ( j -2) of the gain, with this g _n ( j -2) was retained from the normalization processing of the signal vector x _n ( j -1) as the basis for the new normalization gain. From the resulting normalized signal vector x _n ( j ) in step or step 42, using equation (5), the maximum x _{n ,max} absolute values are obtained:

x _{n , max} =max _{0≤ l < L} │ g _n ( j -2) x _{n , l} ( j )│ (5)

In step or step 43, temporal smoothing is applied to x _{n ,max} using a recursive filter taking the previous value x _{n ,max,s m} ( j -2) of said smoothed maximum and resulting in the current temporally smoothed maximum x _{n , max,s m} ( j -1). The purpose of this smoothing is to reduce the adaptation of the normalization gain over time, which reduces the number of gain changes and therefore the amplitude modulation of the signal. Time smoothing is only applied if x _{n, max} is within the range of predefined values. Otherwise, x _{n ,max,s m} ( j -1) is set to x _{n, max} (i.e. the value of x _{n, max} is kept as it is), because subsequent processing must weaken the effective value of x _{n, max} up to the range of preset values. Therefore, time smoothing is only enabled when the normalization gain is constant or when the signal x _n (j ) can be amplified without leaving the range of values.

( x _{n ,max,s m} ( j -1)) is calculated in step/step 43 as follows:

(6)

(6)

where 0 < a ≤ 1 is the attenuation constant.

In order to reduce the bit rate for transmitting the vector e , the normalization gain is calculated from the current time-smoothed maximum value x _{n ,max,s m} ( j -1) and transmitted as an exponent in base '2'. Thus, the following condition must be met:

(7)

(7)

and the quantized exponent e _n ( j -1) is obtained from

(8)

(eight)

at step or step 44.

In periods in which the signal is re-amplified (i.e., the value of the overall gain increases with time), in order to use the available resolution for efficient PCM coding, the exponent e _n ( j ) can be limited (and thus the gain difference between successive blocks) with a small maximum value, such as '1'. This action has two advantageous technical results. On the one hand, small gain differences between successive blocks result in only small amplitude modulations across the transfer function, resulting in reduced crosstalk between adjacent subbands of the FFT spectrum (see the related description of the effect of the transfer function on perceptual coding in conjunction with FIG. 7). ). On the other hand, the bit rate for encoding the exponent is reduced by limiting the range of its values.

Total maximum gain value

(9)

(9)

can be limited to eg '1'. The reason is that if one of the coefficient signals exhibits a large amplitude change between two successive blocks, of which the first has very small amplitudes and the second has the maximum possible amplitude (normalization of the HOA representation in the spatial domain is assumed), then very large differences gain between these two blocks will result in large amplitude modulations through the transfer function, resulting in severe crosstalk between adjacent subbands of the FFT spectrum. This might not be optimal for the subsequent perceptual encoding discussed below.

In step or step 45, the value of the exponent e _n ( j -1) is applied to the transfer function to obtain the current value g _n ( j -1) of the gain. For a continuous transition from the value of g _n ( j -2) of the gain to the value of g _n ( j -1) of the gain, the function shown in Fig.5 is used. The computational rule for this function is as follows:

(10)

(ten)

where l =0, 1, 2, ..., L -1 . The actual vector h _n ( j -1)=[ h _n (0), ..., h _n ( L -1)] ^T of the transfer function with

(11)

(eleven)

будет приводить в результате к требуемому усилению g _n ( j -1) для нормирования x _n ( j ) из уравнения (9).used for continuous attenuation from g _n ( j -2) to g _n ( j -1) . For each value of e _n ( j -1) the value of h _n (0) is equal to e _n ( j -2) since f (0)=1 . The last value of f (L-1) is 0.5 , so

will result in the required gain g _n ( j -1) to normalize x _n ( j ) from equation (9).

In step or step 46, the discrete samples of the vector x _n ( j -1) of the signal are weighted by the values of the gain vector h _n ( j -1) of the transfer function to obtain

x' _n ( j -1)= x _n ( j -1)⊗ h _n ( j -one),(12)

where the operator " ⊗ " is a vector elementwise multiplication of two vectors. This multiplication can also be considered as representing the amplitude modulation of the signal x _n ( j -1) .

In more detail, the coefficients of the vector h _n ( j -1)=[ h _n (0), ..., h _n ( L -1)] ^T of the transfer function are multiplied by the corresponding coefficients of the vector x _n ( j -1) of the signal, where the value of h _n (0) is h _n (0) = g _n ( j -2) , and the value of h _n ( L -1) is h _n ( L -1)= g _n ( j -1) . Therefore, the transfer function is continuously attenuated from the gain value g _n ( j -2) to the gain value g _n ( j -1) as shown in the example of Fig. 8, which shows the gain values from the transfer functions h _n ( j ) , h _n ( j -1) and h _n ( j -2) which are applied to the respective signal vectors x _n ( j ) , x _n ( j -1) and x _n ( j -2) for three consecutive blocks. The advantage over subsequent perceptual encoding is that at block boundaries the applied gains are continuous: the transfer function h _n ( j -1) continuously attenuates the gains for the coefficients x _n ( j -1) with g _n ( j -2) to g _n ( j -1) .

The adaptive denormalization processing at the decoder or receiver side is shown in FIG. Input values are PCM-encoded and normalized signals x” _n ( j -one), a suitable indicator e _n ( j -one) degrees and meaning g _n ( j -2) gain of the last block. Meaning g _n ( j -2) gain of the last block is calculated recursively, and g _n ( j -2) must be initialized with a predefined value that was also already used in the encoder. The output results are the value g _n ( j -one)gain from step/step 61 and the denormalized signal x''' _n ( j -one) from step/step 62.

In step or step 61, the exponent is applied to the transfer function. To restore the range of values x _n ( j -1) , in equation (11) the transfer function vector h _n ( j -1) is calculated from the received exponent e _n ( j -1) of the degree and the recursively calculated gain g _n ( j -2 ) . The gain g _n ( j -1) for processing the next block is set to h _n ( L -1) .

In step or step 62, an inverse gain is applied. The applied amplitude modulation of normalization processing is inverted by x''' _n ( j -1)= x” _n ( j -1)⊗ h _n ( j -1) ^-1 , (13)

, а «⊗» является векторным поэлементным умножением, которое было использовано на стороне кодера или передатчика. Дискретные отсчеты x’ _n ( j -1) не могут быть представлены входным PCM-форматом x” _n ( j -1), так что денормированию требуется преобразование в формат диапазона больших значений, как например формат с плавающей запятой.where

, and " ⊗ " is the elementwise vector multiplication that was used on the encoder or transmitter side. Discrete samples x' _n ( j -1) cannot be represented by the input PCM format x” _n ( j -1), so denormalization needs to be converted to a large range format, such as floating point format.

With regard to the transmission of ancillary information, for the transmission of exponents e _n ( j -1) of the degree, their probability cannot be assumed to be uniform, because the applied normalization gain will be constant for successive blocks of the same range of values. Thus, entropy encoding, such as Huffman encoding, can be applied to exponent values to reduce the required data rate.

One disadvantage of the described processing may be the recursive calculation of the gain value g _n ( j -2) . Therefore, denormalization processing can only start from the beginning of the HOA stream.

The solution to this problem is to add access blocks to the HOA format to provide information for regular computation. g _n ( j -2). In this case, the access block must provide indicators e _{n, access} =log ₂ g _n ( j -2) (fourteen)

могло быть вычислено и денормирование могло начаться в каждом t -ом блоке.degree for each t -th block so that

could be computed and denormalization could start at every tth block.

(15)The impact on the perceptual coding of the normalized signal x' _n ( j -1) is analyzed by the absolute value of the frequency response

(fifteen)

functions h _n ( l ) . The frequency response is given by a Fast Fourier Transform (FFT) over h _n ( l ) as shown in equation (15).

Figure 7 shows the normalized (to 0 dB) amplitude FFT spectrum H _n ( u ) to clarify the spectral distortion introduced by amplitude modulation. Attenuation │ H _n ( u )│ is relatively sharp for small exponents and becomes flat for larger exponents.

Since amplitude modulation x _n ( j -1) over h _n ( l ) in the time domain is equivalent to convolution over H _n ( u ) in the frequency domain, the abrupt decay of the frequency response of H _n ( u ) reduces crosstalk between adjacent FFT-subbands. spectrum x' _n ( j -1). This is strongly related to the subsequent perceptual encoding x' _n ( j -1) because subband crosstalk affects the estimated perceptual characteristics of the signal. Thus, for abrupt fading H _n ( u ) , the perceptual encoding assumptions for x' _n ( j -1) are also valid for the non-normalized signal x _n ( j -1) .

This shows that for small exponents the perceptual encoding x' _n ( j -1) is almost equivalent to the perceptual encoding x _n ( j -1) and that the perceptual encoding of the normalized signal has almost no effect on the denormalized signal at low exponent amplitude.

The claimed processing may be performed by a single processor or electronic circuit on the transmitting side and on the receiving side, or by several processors or electronic circuits operating in parallel and/or operating in different parts of the claimed processing.

Claims (13)

1. A method for decoding the representation of HOA (High Order Ambisonics) signals in the coefficient domain, said decoding comprising: - transforming the vector of PCM-coded signals in the spatial domain from the representation of the HOA signals in the coefficient domain into the corresponding vector of signals in the coefficient domain by multiplying the vector of PCM-coded signals in the spatial domain by the transformation matrix; - denormalization of a vector of PCM-encoded and normalized coefficient-domain signals from the coefficient-domain representation of HOA signals, said denormalization comprising: -- determining a transfer function vector based on a corresponding ancillary information exponent and a recursively computed gain value, wherein the corresponding exponent and gain value are based on a variable index of the input matrix of HOA signal vectors; -- applying the appropriate inverse gain value to the vector of PCM-encoded and normalized signals in the coefficient domain in order to determine the corresponding vector of PCM-encoded and denormalized signal; combining the vector of signals in the coefficient domain and the vector of denormalized signals in the coefficient domain to determine the combined vector of HOA signals in the coefficient domain, which may have a variable number of HOA coefficients. 2. A device for decoding the representation of HOA (High Order Ambisonics) signals in the coefficient domain, said decoding device comprising: means for converting a vector of PCM encoded signals in the spatial domain from a coefficient domain representation of HOA signals to a corresponding vector of signals in the coefficient domain by multiplying the vector of PCM encoded signals in the spatial domain by a transformation matrix; means for denormalizing said vector of PCM-encoded and normalized signals in the coefficient domain, including: -- means configured to determine a transfer function vector based on a corresponding ancillary information exponent and a recursively computed gain value, wherein the corresponding exponent and gain value are based on a variable index of the input matrix of HOA signal vectors; -- means configured to apply a corresponding reciprocal gain value to a vector of PCM-encoded and normalized signals in the coefficient domain in order to determine a corresponding vector of PCM-encoded and denormalized signal; and - means configured to combine the vector of signals in the coefficient domain and the vector of denormalized signals in the coefficient domain to determine the combined vector of HOA signals in the coefficient domain, which may have a variable number of HOA coefficients. 3. A storage medium that stores executable instructions which, when executed, cause the processor to execute the method of claim 1.

Images