| Full Text |
Description
METHOD AND APPARATUS FOR PROCESSING A MEDIA
SIGNAL
Technical Field
[ 1 ] The present invention relates to an apparatus for processing a media signal and
method thereof, and more particularly to an apparatus for generating a surround signal
by using spatial information of the media signal and method thereof.
Background Art
[2] Generally, various kinds of apparatuses and methods have been widely used to
generate a multi-channel media signal by using spatial information for the multichannel
media signal and a downmix signal, in which the downmix signal is generated
by downmixing the multi-channel media signal into mono or stereo signal.
[3] However, the above methods and apparatuses are not usable in environments
unsuitable for generating a multi-channel signal. For instance, they are not usable for a
device capable of generating only a stereo signal. In other words, there exists no
method or apparatus for generating a surround signal, in which the surround signal has
multi-channel features in the environment incapable of generating a multi-channel
signal by using spatial information of the multi-channel signal.
[4] So, since there exists no method or apparatus for generating a surround signal in a
device capable of generating only a mono or stereo signal, it is difficult to process the
media signal efficiently.
Disclosure of Invention
Technical Problem
[5] Accordingly, the present invention is directed to an apparatus for processing a
media signal and method thereof that substantially obviate one or more of the problems
due to limitations and disadvantages of the related art.
[6] An object of the present invention is to provide an apparatus for processing a media
signal and method thereof, by which the media signal can be converted to a surround
signal by using spatial information for the media signal.
[7] Additional features and advantages of the invention will be set forth in a description
which follows, and in part will be apparent from the description, or may be learned by
practice of the invention. The objectives and other advantages of the invention will be
realized and attained by the structure particularly pointed out in the written description
and claims thereof as well as the appended drawings.
Technical Solution
[8] To achieve these and other advantages and in accordance with the purpose of the
present invention, a method of processing a signal according to the present invention
includes of: generating source mapping information corresponding to each source of
multi-sources by using spatial information indicating features between the multi-
sources; generating sub-rendering information by applying filter information giving a
surround effect to the source mapping information per the source; generating rendering
information for generating a surround signal by integrating at least one of the sub-
rendering information; and generating the surround signal by applying the rendering
information to a downmix signal generated by downmixing the multi-sources.
[9] To further achieve these and other advantages and in accordance with the purpose .
of the present invention, an apparatus for processing a signal includes a source map
ping unit generating source mapping information corresponding to each source of
multi-sources by using spatial information indicating features between the multi-
sources; a sub-rendering information generating unit generating sub-rendering in-
formation by applying filter information having a surround effect to the source
mapping information per the source; an integrating unit generating rendering in-
formation for generating a surround signal by integrating the at least one of the sub-
rendering information; and a rendering unit generating the surround signal by applying
the rendering information to a downmix signal generated by downmixing the multi-
sources.
[10] It is to be understood that both the foregoing general description and the following
detailed description are exemplary and explanatory and are intended to provide further
explanation of the invention as claimed.
Advantageous Effects
[11] A signal processing apparatus and method according to the present invention enable
a decoder, which receives a bitstream including a downmix signal generated by
downmixing a multi-channel signal and spatial information of the multi-channel signal,
to generate a signal having a surround effect in environments in incapable of
recovering the multi-channel signal.
Brief Description of the Drawings
[12] The accompanying drawings, which are included to provide a further understanding
of the invention and are incorporated in and constitute a part of this specification,
illustrate embodiments of the invention and together with the description serve to
explain the principles of the invention.
[13] In the drawings:
[14] FIG. 1 is a block diagram of an audio signal encoding apparatus and an audio signal
decoding apparatus according to one embodiment of the present invention;
[15] FIG. 2 is a structural diagram of a bitstream of an audio signal according to one
embodiment of the present invention;
[16] FIG. 3 is a detailed block diagram of a spatial information converting unit
according to one embodiment of the present invention;
[17] FIG. 4 and FIG. 5 are block diagrams of channel configurations used for source
mapping process according to one embodiment of the present invention;
[18] FIG. 6 and FIG. 7 are detailed block diagrams of a rendering unit for a stereo
downmix signal according to one embodiment of the present invention;
[19] FIG. 8 and FIG. 9 are detailed block diagrams of a rendering unit for a mono
downmix signal according to one embodiment of the present invention;
[20] FIG. 10 and FIG. 11 are block diagrams of a smoothing unit and an expanding unit
according to one embodiment of the present invention;
[21] FIG. 12 is a graph to explain a first smoothing method according to one
embodiment of the present invention;
[22] FIG. 13 is a graph to explain a second smoothing method according to one
embodiment of the present invention;
[23] FIG. 14 is a graph to explain a third smoothing method according to one
embodiment of the present invention;
[24] FIG. 15 is a graph to explain a fourth smoothing method according to one
embodiment of the present invention;
[25] FIG. 16 is a graph to explain a fifth smoothing method according to one
embodiment of the present invention;
[26] FIG. 17 is a diagram to explain prototype filter information corresponding to each
channel;
[27] FIG. 18 is a block diagram for a first method of generating rendering filter in-
formation in a spatial information converting unit according to one embodiment of the
present invention;
[28] FIG. 19 is a block diagram for a second method of generating rendering filter in-
formation in a spatial information converting unit according to one embodiment of the
present invention;
[29] FIG. 20 is a block diagram for a third method of generating rendering filter in-
formation in a spatial information converting unit according to one embodiment of the
present invention;
[30] FIG. 21 is a diagram to explain a method of generating a surround signal in a
rendering unit according to one embodiment of the present invention;
[31] FIG. 22 is a diagram for a first interpolating method according to one embodiment
of the present invention;
[32] FIG. 23 is a diagram for a second interpolating method according to one
embodiment of the present invention;
[33] FIG. 24 is a diagram for a block switching method according to one embodiment of
the present invention;
[34] FIG. 25 is a block diagram for a position to which a window length decided by a
window, length deciding unit is applied according to one embodiment of the present
invention;
[35] FIG. 26 is a diagram for filters having various lengths used in processing an audio
signal according to one embodiment of the present invention;
[36] FIG. 27 is a diagram for a method of processing an audio signal dividedly by using
a plurality of subfilters according to one embodiment of the present invention;
[37] FIG. 28 is a block diagram for a method of rendering partition rendering in-
formation generated by a plurality of subfilters to a mono downmix signal according to
one embodiment of the present invention;
[38] FIG. 29 is a block diagram for a method of rendering partition rendering in-
formation generated by a plurality of subfilters to a stereo downmix signal according to
one embodiment of the present invention;
[39] FIG. 30 is a block diagram for a first domain converting method of a downmix
signal according to one embodiment of the present invention; and
[40] FIG. 31 is a block diagram for a second domain converting method of a downmix
signal according to one embodiment of the present invention.
Best Mode for Carrying Out the Invention
[41] Reference will now be made in detail to the preferred embodiments of the present
invention, examples of which are illustrated in the accompanying drawings..
[42] FIG. 1 is a block diagram of an audio signal encoding apparatus and an audio signal
decoding apparatus according to one embodiment of the present invention.
[43] Referring to FIG. 1, an encoding apparatus 10 includes a downmixing unit 100, a
spatial information generating unit 200, a downmix signal encoding unit 300, a spatial
information encoding unit 400, and a multiplexing unit 500.
[44] If multi-source(Xl, X2,..., Xn) audio signal is inputted to the downmixing unit 100,
the downmixing unit 100 downmixes the inputted signal into a downmix signal. In this
case, the downmix signal includes mono, stereo and multi-source audio signal.
[45] The source includes a channel and ,in convenience, is represented as a channel in
the following description. In the present specification, the mono or stereo downmix
signal is referred to as a reference. Yet, the present invention is not limited to the mono
or stereo downmix signal.
[46] The encoding apparatus 10 is able to optionally use an arbitrary downmix signal
directly provided from an external environment.
[47] The spatial information generating unit 200 generates spatial information from a
multi-channel audio signal. The spatial information can be generated in the course of a
downmixing process. The generated downmix signal and spatial information are
encoded by the downmix signal encoding unit 300 and the spatial information
encoding unit 400, respectively and are then transferred to the multiplexing unit 500.
[48] In the present invention, 'spatial information' means information necessary to
generate a multi-channel signal from upmixing a downmix signal by a decoding
apparatus, in which the downmix signal is generated by downmixing the multi-channel
signal by an encoding apparatus and transferred to the decoding apparatus. The spatial
information includes spatial parameters. The spatial parameters include CLD(channel
level difference) indicating an energy difference between channels, ICC(inter-channel
coherences) indicating a correlation between channels, CPC(channel prediction co-
efficients) used in generating three channels from two channels, etc.
[49] In the present invention, 'downmix signal encoding unit' or 'downmix signal
decoding unit' means a codec that encodes or decodes an audio signal instead of spatial
information. In the present specification, a downmix audio signal is taken as an
example of the audio signal instead of the spatial information. And, the downmix
signal encoding or decoding unit may include MP3, AC-3, DTS, or AAC. Moreover,
the downmix signal encoding or decoding unit may include a codec of the future as
well as the previously developed codec.
[50] The multiplexing unit 500 generates a bitstream by multiplexing the downmix
signal and the spatial information and then transfers the generated bitstream to the
decoding apparatus 20. Besides, the structure of the bitstream will be explained in FIG.
2 later.
[51] A decoding apparatus 20 includes a demultiplexing unit 600, a downmix signal
decoding unit 700, a spatial information decoding unit 800, a rendering unit 900, and a
spatial information converting unit 1000.
[52] The demultiplexing unit 600 receives a bitstream and then separates an encoded
downmix signal and an encoded spatial information from the bitstream. Subsequently,
the downmix signal decoding unit 700 decodes the encoded downmix signal and the
spatial information decoding unit 800 decodes the encoded spatial information.
[53] The spatial information converting unit 1000 generates rendering information
applicable to a downmix signal using the decoded spatial information and filter in-
formation. In this case, the rendering information is applied to the downmix signal to
generate a surround signal.
[54] For instance, the surround signal is generated in the following manner. First of all, a
process for generating a downmix signal from a multi-channel audio signal by the
encoding apparatus 10 can include several steps using an OTT (one-to-two) or TTT
(three-to-three) box. In this case, spatial information can be generated from each of the
steps. The spatial information is transferred to the decoding apparatus 20. The
decoding apparatus 20 then generates a surround signal by converting the spatial in-
formation and then rendering the converted spatial information with a downmix signal.
Instead of generating a multi-channel signal by upmixing a downmix signal, the
present invention relates to a rendering method including the steps of extracting spatial
information for each upmixing step and performing a rendering by using the extracted
spatial information. For example, HRTF (head-related transfer functions) filtering is
usable in the rendering method.
[55] In this case, the spatial information is a value applicable to a hybrid domain as well.
So, the rendering can be classified into the following types according to a domain.
[56] The first type is that the rendering is executed on a hybrid domain by having a
downmix signal pass through a hybrid filterbank. In this case, a conversion of domain
for spatial information is unnecessary.
[57] The second type is that the rendering is executed on a time domain. In this case, the
second type uses a fact that a HRTF filter is modeled as a FIR (finite inverse response)
filter or an IIR (infinite inverse response) filter on a time domain. So, a process for
converting spatial information to a filter coefficient of time domain is needed.
[58] The third type is that the rendering is executed on a different frequency domain. For
instance, the rendering is executed on a DFT (discrete Fourier transform) domain. In
this case, a process for transforming spatial information into a corresponding domain is
necessary. In particular, the third type enables a fast operation by replacing a filtering
on a time domain into an operation on a frequency domain.
[59] In the present invention, filter information is the information for a filter necessary
for processing an audio signal and includes a filter coefficient provided to a specific
filter. Examples of the filter information are explained as follows. First of all,
prototype filter information is original filter information of a specific filter and can be
represented as GL_L or the like. Converted filter information indicates a filter co-
efficient after the prototype filter information has been converted and can be
represented as GL_L or the like. Sub-rendering information means the filter in-
formation resulting from spatializing the prototype filter information to generate a
surround signal and can be represented as FL_L1 or the like. Rendering information
means the filter information necessary for executing rendering and can be represented
as HLJL or the like. Interpolated/smoothed rendering information means the filter in-
formation resulting from interpolation/smoothing the rendering information and can be
represented as HL_L or the like. In the present specification, the above filter in-
formations are referred to. Yet, the present invention is not restricted by the names of
the filter informations. In particular, HRTF is taken as an example of the filter in-
formation. Yet, the present invention is not limited to the HRTF.
[60] The rendering unit 900 receives the decoded downmix signal and the rendering in-
formation and then generates a surround signal using the decoded downmix signal and
the rendering information. The surround signal may be the signal for providing a
surround effect to an audio system capable of generating only a stereo signal. Besides,
the present invention can be applied to various systems as well as the audio system
capable of generating only the stereo signal.
[61] FIG. 2 is a structural diagram for a bitstream of an audio signal according to one
embodiment of the present invention, in which the bitstream includes an encoded
downmix signal and encoded spatial information.
[62] Referring to FIG. 2, a 1-frame audio payload includes a downmix signal field and
an ancillary data field. Encoded spatial information can be stored in the ancillary data
field. For instance, if an audio payload is 48~128kbps, spatial information can have a
range of 5~32kbps. Yet, no limitations are put on the ranges of the audio payload and
spatial information.
[63] FIG. 3 is a detailed block diagram of a spatial information converting unit
according to one embodiment of the present invention.
[64] Referring to FIG. 3, a spatial information converting unit 1000 includes a source
mapping unit 1010, a sub-rendering information generating unit 1020, an integrating
unit 1030, a processing unit 1040, and a domain converting unit 1050.
[65] The source mapping unit 101 generates source mapping information corresponding
to each source of an audio signal by executing source mapping using spatial in-
formation. In this case, the source mapping information means per-source information
generated to correspond to each source of an audio signal by using spatial information
and the like. The source includes a channel and ,in this case, the source mapping in-
formation corresponding to each channel is generated. The source mapping in-
formation can be represented as a coefficient. And, the source mapping process will be
explained in detail later with reference to FIG. 4 and FIG. 5.
[66] The sub-rendering information generating unit 1020 generates sub-rendering in-
formation corresponding to each source by using the source mapping information and
the filter information. For instance, if the rendering unit 900 is the HRTF filter, the
sub-rendering information generating unit 1020 is able to generate sub-rendering in-
formation by using HRTF filter information.
[67] The integrating unit 1030 generates rendering information by integrating the sub-
rendering information to correspond to each source of a downmix signal. The
rendering information, which is generated by using the spatial information and the
filter information, means the information to generate a surround signal by being
applied to the downmix signal. And, the rendering information includes a filter co-
efficient type. The integration can be omitted to reduce an operation quantity of the
rendering process. Subsequently, the rendering information is transferred to the
processing unit 1042.
[68] The processing unit 1042 includes an interpolating unit 1041 and/or a smoothing
unit 1042. The rendering information is interpolated by the interpolating unit 1041
and/or smoothed by the smoothing unit 1042.
[69] The domain converting unit 1050 converts a domain of the rendering information to
a domain of the downmix signal used by the rendering unit 900. And, the domain
converting unit 1050 can be provided to one of various positions including the position
shown in FIG. 3. So, if the rendering information is generated on the same domain of
the rendering unit 900, it is able to omit the domain converting unit 1050. The domain-
converted rendering information is then transferred to the rendering unit 900.
[70] The spatial information converting unit 1000 can include a filter information
converting unit 1060. In FIG. 3, the filter information converting unit 1060 is provided
within the spatial information converting unit 100. Alternatively, the filter information
converting unit 1060 can be provided outside the spatial information converting unit
100. The filter information converting unit 1060 is converted to be suitable for
generating sub-rendering information or rendering information from random filter in-
formation, e.g., HRTF. The converting process of the filter information can include the
following steps.
[71] First of all, a step of matching a domain to be applicable is included. If a domain of
filter information does not match a domain for executing rendering, the domain
matching step is required. For instance, a step of converting time domain HRTF to
DFT, QMF or hybrid domain for generating rendering information is necessary.
[72] Secondly, a coefficient reducing step can be included. In this case, it is easy to save
the domain-converted HRTF and apply the domain-converted HRTF to spatial in-
formation. For instance, if a prototype filter coefficient has a response of a long tap
number (length), a corresponding coefficient has to be stored in a memory cor-
responding to a response amounting to a corresponding length of total 10 in case of 5.1
channels. This increases a load of the memory and an operational quantity. To prevent
this problem, a method of reducing a filter coefficient to be stored while maintaining
filter characteristics in the domain converting process can be used. For instance, the
HRTF response can be converted to a few parameter value. In this case, a parameter
generating process and a parameter value can differ according to an applied domain.
[73] The downmix signal passes through a domain converting unit 1110 and/or a
decorrelating unit 1200 before being rendered with the rendering information. In case
that a domain of the rendering information is different from that of the downmix
signal, the domain converting unit 1110 converts the domain of the downmix signal in
order to match the two domains together.
[74] The decorrelating unit 1200 is applied to the domain-converted downmix signal.
This may have an operational quantity relatively higher than that of a method of
applying a decorrelator to the rendering information. Yet, it is able to prevent
distortions from occurring in the process of generating rendering information. The
decorrelating unit 1200 can include a plurality of decorrelators differing from each
other in characteristics if an operational quantity is allowable. If the downmix signal is
a stereo signal, the decorrelating unit 1200 may not be used. In FIG. 3, in case that a
domain-converted mono downmix signal, i.e., a mono downmix signal on a frequency,
hybrid, QMF or DFT domain is used in the rendering process, a decorrelator is used on
the corresponding domain. And, the present invention includes a decorrelator used on a
time domain as well. In this case, a mono downmix signal before the domain
converting unit 1100 is directly inputted to the decorrelating unit 1200. A first order or
higher IIR filter (or FIR filter) is usable as the decorrelator.
[75] Subsequently, the rendering unit 900 generates a surround signal using the
downmix signal, the decorrelated downmix signal, and the rendering information. If
the downmix signal is a stereo signal, the decorrelated downmix signal may not be
used. Details of the rendering process will be described later with reference to FIGs. 6
to 9.
[76] The surround signal is converted to a time domain by an inverse domain converting
unit 1300 and then outputted. If so, a user is able to listen to a sound having a multi-
channel effect though stereophonic earphones or the like.
[77] FIG. 4 and FIG. 5 are block diagrams of channel configurations used for source
mapping process according to one embodiment of the present invention. A source
mapping process is a process for generating source mapping information cor-
responding to each source of an audio signal by using spatial information. As
mentioned in the foregoing description, the source includes a channel and source
mapping information can be generated to correspond to the channels shown in FIG. 4
and FIG. 5. The source mapping information is generated in a type suitable for a
rendering process.
[78] For instance, if a downmix signal is a mono signal, it is able to generate source
mapping information using spatial information such as CLD1~CLD5, ICC1~ICC5, and
the like.
[79] The source mapping information can be represented as such a value as D_L (=DL ),
D_R (=DR ), D_C (=DC ), D_LFE (=DLFE), D_Ls (=DLS), D_Rs (=DRS), and the like. In
R C LFE Ls Rs
this case, the process for generating the source mapping information is variable
according to a tree structure corresponding to spatial information, a range of spatial in-
formation to be used, and the like. In the present specification, the downmix signal is a
mono signal for example, which does not put limitation of the present invention.
[80] Right and left channel outputs outputted from the rendering unit 900 can be
expressed as Math Figure 1.
[81] MathFigure 1
[82] In this case, the operator '*' indicates a product on a DFT domain and can be
replaced by a convolution on a QMF or time domain.
[83] The present invention includes a method of generating the L, C, R, Ls and Rs by
source mapping information using spatial information or by source mapping in-
formation using spatial information and filter information. For instance, source
mapping information can be generated using CLD of spatial information only or CLD
and ICC of spatial information. The method of generating source mapping information
using the CLD only is explained as follows.
[84] In case that the tree structure has a structure shown in FIG. 4, a first method of
obtaining source mapping information using CLD only can be expressed as Math
Figure 2.
[85] MathFigure 2
, and 'm' indicates a mono downmix signal.
[87] In case that the tree structure has a structure shown in FIG. 5, a second method of
obtaining source mapping information using CLD only can be expressed as Math
Figure 3.
[88] MathFigure 3
[89] If source mapping information is generated using CLD only, a 3-dimensional effect
may be reduced. So, it is able to generate source mapping information using ICC and/
or decorrelator. And, a multi-channel signal generated by using a decorrelator output
signal dx(m) can be expresses as Math Figure 4.
[90] MathFigure 4
[91] In this case, 'A', 'B' and 'C are values that can be represented by using CLD and
ICC. 'd0' to 'd3' indicate decorrelators. And, 'm' indicates a mono downmix signal. Yet,
this method is unable to generate source mapping information such as D_L, D_R, and
the like.
[92] Hence, the first method of generating the source mapping information using the
CLD, ICC and/or decorrelators for the downmix signal regards dx(m) (x=0, 1, 2) as an
independent input. In this case, the 'dx' is usable for a process for generating sub-
rendering filter information according to Math Figure 5.
[93] MathFigure 5
FL_L_M = d_L_M*GL_L' (Mono input Left output)
FL_R_M = d_L_M*GL_R' (Mono input Right output)
FL_L_Dx = d_L_Dx*GL_L' (Dx output Left output)
FL_R_Dx = d_L_Dx*GL_R' (Dx output Right output)
[94] And, rendering information can be generated according to Math Figure 6 using a
result of Math Figure 5.
[95] MathFigure 6
HM_L = FL_L_M + FR_L_M + FC_L_M + FLS_L_M + FRS_L_M + FLFE_L_M
HM_R = FL_R_M + FR_R_M + FC_R_M + FLS_R_M + FRS_R_M + FLFE_R_M
HDx_L = FL_L_Dx + FR_L_Dx + FC_L_Dx + FLS_L_Dx + FRS_L_Dx + FLFE_L_Dx
HDx_R = FL_R_Dx + FR_R_Dx + FC_R_Dx + FLS_R_Dx + FRS_R_Dx + FLFE_R_Dx
[96] Details of the rendering information generating process are explained later. The first
method of generating the source mapping information using the CLD, ICC and/or
decorrelators handles a dx output value, i.e., 'dx(m)' as an independent input, which
may increase an operational quantity.
[97] A second method of generating source mapping information using CLD, ICC and/
or decorrelators employs decorrelators applied on a frequency domain. In this case, the
source mapping information can be expresses as Math Figure 7.
[98] MathFigure 7
[99] In this case, by applying decorrelators on a frequency domain, the same source
mapping information such as D_L, D_R, and the like before the application of the
decorrelators can be generated. So, it can be implemented in a simple manner.
[100] A third method of generating source mapping information using CLD, ICC and/or
decorrelators employs decorrelators having the all-pass characteristic as the
decorrelators of the second method. In this case, the all-pass characteristic means that a
size is fixed with a phase variation only. And, the present invention can use
decorrelators having the all-pass characteristic as the decorrelators of the first method.
[101] A fourth method of generating source mapping information using CLD, ICC and/or
decorrelators carries out decorrelation by using decorrelators for the respective
channels (e.g., L, R, C, Ls, Rs, etc.) instead of using 'd0' to 'd3' of the second method. In
this case, the source mapping information can be expressed as Math Figure 8.
[102] MathFigure8
[103] In this case, 'k' is an energy value of a decorrelated signal determined from CLD
and ICC values. And, 'd_L', 'd_R', 'd_C, 'd_Ls' and 'd_Rs' indicate decorrelators
applied to channels, respectively.
[104] A fifth method of generating source mapping information using CLD, ICC and/or
decorrelators maximizes a decorrelation effect by configuring 'd_L' and 'd_R'
symmetric to each other in the fourth method and configuring 'd_Ls' and 'd_Rs'
symmetric to each other in the fourth method. In particular, assuming d_R=f(d_L) and
d_Rs=f(d_Ls), it is necessary to design 'd_L', 'd_C and 'd_Ls' only.
. [105] A sixth method of generating source mapping information using CLD, ICC and/or
decorrelators is to configure the 'd_L' and 'd_Ls' to have a correlation in the fifth
method. And, the 'd_L' and 'd_C can be configured to have a correlation as well.
[106] A seventh method of generating source mapping information using CLD, ICC and/
or decorrelators is to use the decorrelators in the third method as a serial or nested
structure of the all-pas filters. The seventh method utilizes a fact that the all-pass char-
acteristic is maintained even if the all-pass filter is used as the serial or nested
structure. In case of using the all-pass filter as the serial or nested structure, it is able to
obtain more various kinds of phase responses. Hence, the decorrelation effect can be
maximized.
[107] An eighth method of generating source mapping information using CLD, ICC and/
or decorrelators is to use the related art decorrelator and the frequency-domain
decorrelator of the second method together. In this case, a multi-channel signal can be
expressed as Math Figure 9.
[108] MathFigure 9
[109] In this case, a filter coefficient generating process uses the same process explained
in the first method except that 'A' is changed into 'A+Kd'.
[110] A ninth method of generating source mapping information using CLD, ICC and/or
decorrelators is to generate an additionally decorrelated value by applying a frequency
domain decorrelator to an output of the related art decorrelator in case of using the
related art decorrelator. Hence, it is able to generate source mapping information with
a small operational quantity by overcoming the limitation of the frequency domain
decorrelator.
[Ill] A tenth method of generating source mapping information using CLD, ICC and/or
decorrelators is expressed as Math Figure 10.
[112] MathFigure 10
[113] In this case, 'di_(m)'(i=L, R, C, Ls, Rs) is a decorrelator output value applied to a
channel-i. And, the output value can be processed on a time domain, a frequency
domain, a QMF domain, a hybrid domain, or the like. If the output value is processed
on a domain different from a currently processed domain, it can be converted by
domain conversion. It is able to use the same 'd for d_L, d_R, d_C, d_Ls, and d_Rs. In
this case, Math Figure 10 can be expressed in a very simple manner.
[114] If Math Figure 10 is applied to Math Figure 1, Math Figure 1 can be expressed as
Math Figure 11.
[115] MathFigure ll
Lo = HM_L*m + HMD_L*d(m)
Ro = HM_R*R + HMD_R*d(m)
[116] In this case, rendering information HMJL is a value resulting from combining
spatial information and filter information to generate a surround signal Lo with an
input m. And, rendering information HM_R is a value resulting from combining spatial
information and filter information to generate a surround signal Ro with an input m.
Moreover, 'd(m)' is a decorrelator output value generated by transferring a decorrelator
output value on an arbitrary domain to a value on a current domain or a decorrelator
output value generated by being processed on a current domain. Rendering information
HMD_L is a value indicating an extent of the decorrelator output value d(m) that is
added to 'Lo' in rendering the d(m), and also a value resulting from combining spatial
information and filter information together. Rendering information HMD_R is a value
indicating an extent of the decorrelator output value d(m) that is added to 'Ro' in
rendering the d(m).
[117] Thus, in order to perform a rendering process on a mono downmix signal, the
present invention proposes a method of generating a surround signal by rendering the
rendering information generated by combining spatial information and filter in-
formation (e.g., HRTF filter coefficient) to a downmix signal and a decorrelated
downmix signal. The rendering process can be executed regardless of domains. If
'd(m)' is expressed as 'd*m'(product operator) being executed on a frequency domain,
Math Figure 11 can be expressed as Math Figure 12.
[118] MathFigure l2
Lo = HH_L*m + HMD_L*d*m = HMoveralI_L*m
Ro = HM_R*m + HMD_R*d*m = HMoverall_R*m
[119] Thus, in case of performing a rendering process on a downmix signal on a
frequency domain, it is ale to minimize an operational quantity in a manner of rep-
resenting a value resulting from combining spatial information, filter information and
decorrelators appropriately as a product form.
[120] FIG. 6 and FIG. 7 are detailed block diagrams of a rendering unit for a stereo
downmix signal according to one embodiment of the present invention.
[121] Referring to FIG. 6, the rendering unit 900 includes a rendering unit-A 910 and a
rendering unit-B 920.
[122] If a downmix signal is a stereo signal, the spatial information converting unit 1000
generates rendering information for left and right channels of the downmix signal. The
rendering unit-A 910 generates a surround signal by rendering the rendering in-
formation for the left channel of the downmix signal to the left channel of the
downmix signal. And, the rendering unit-B 920 generates a surround signal by
rendering the rendering information for the right channel of the downmix signal to the
right channel of the downmix signal. The names of the channels are just exemplary,
which does not put limitation on the present invention.
[123] The rendering information can include rendering information delivered to a same
channel and rendering information delivered to another channel.
[124] For instance, the spatial information converting unit 1000 is able to generate
rendering information HL_L and HL_R inputted to the rendering unit for the left
channel of the downmix signal, in which rendering information HL_L is delivered to a
left output corresponding to the same channel and the rendering information HL_R is
delivered to a right output corresponding to the another channel. And, the spatial in-
formation converting unit 1000 is able to generate rendering information HR_R and
HR_L inputted to the rendering unit for the right channel of the downmix signal, in
which the rendering information HR_R is delivered to a right output corresponding to
the same channel and the rendering information HRJL is delivered to a left output cor-
responding to the another channel.
[125] Referring to FIG. 7, the rendering unit 900 includes a rendering unit-1A 911, a
rendering unit-2A 912, a rendering unit-1B 921, and a rendering unit-2B 922.
[126] The rendering unit 900 receives a stereo downmix signal and rendering information
from the spatial information converting unit 1000. Subsequently, the rendering unit
900 generates a surround signal by rendering the rendering information to the stereo
downmix signal.
[127] In particular, the rendering unit-1A 911 performs rendering by using rendering in-
formation HL_L delivered to a same channel among rendering information for a left
channel of a downmix signal. The rendering unit-2A 912 performs rendering by using
rendering information HL_R delivered to a another channel among rendering in-
formation for a left channel of a downmix signal. The rendering unit-1B 921 performs
rendering by using rendering information HR_R delivered to a same channel among
rendering information for a right channel of a downmix signal. And, the rendering
unit-2B 922 performs rendering by using rendering information HR_L delivered to
another channel among rendering information for a right channel of a downmix signal.
[128] In the following description, the rendering information delivered to another channel
is named 'cross-rendering information' The cross-rendering information HL_R or
HR_L is applied to a same channel and then added to another channel by an adder. In
this case, the cross-rendering information HL_R and/or HR_L can be zero. If the cross-
rendering information HL_R and/or HR_L is zero, it means that no contribution is
made to the corresponding path.
[129] An example of the surround signal generating method shown in FIG. 6 or FIG. 7 is
explained as follows.
[130] First of all, if a downmix signal is a stereo signal, the downmix signal defined as 'x',
source mapping information generated by using spatial information defined as 'D',
prototype filter information defined as 'G', a multi-channel signal defined as 'p' and a
surround signal defined as 'y1 can be represented by matrixes shown in Math Figure 13.
[131] MathFigure 13
[132] In this case, if the above values are on a frequency domain, they can be developed
as follows.
[133] First of all, the multi-channel signal p, as shown in Math Figure 14, can be
expressed as a product between the source mapping information D generated by using
the spatial information and the downmix signal x.
[134] MathFigure 14
[135] The surround signal y, as shown in Math Figure 15, can be generated by rendering
the prototype filter information G to the multi-channel signal p.
[136] MathFigure 15
y=G.p
[137] In this case, if Math Figure 14 is inserted in the p, it can be generated as Math
Figure 16.
[138] MathFigure 16
y = GDx
[139] In this case, if rendering information H is defined as H=GD, the surround signal y
and the downmix signal x can have a relation of Math Figure 17.
[140] MathFigure 17
[141] Hence, after the rendering information H has been generated by processing the
product between the filter information and the source mapping information, the
downmix signal x is multiplied by the rendering information H to generate the
surround signal y.
[142] According to the definition of the rendering information H, the rendering in-
formation H can be expressed as Math Figure 18.
[143] MathFigure 18
H = GD
[144] FIG. 8 and FIG. 9 are detailed block diagrams of a rendering unit for a mono
downmix signal according to one embodiment of the present invention.
[145] Referring to FIG. 8, the rendering unit 900 includes a rendering unit-A 930 and a
rendering unit-B 940.
[146] If a downmix signal is a mono signal, the spatial information converting unit 1000
generates rendering information HM_L and HM_R, in which the rendering in-
formation HM_L is used in rendering the mono signal to a left channel and the
rendering information HM_R is used in rendering the mono signal to a right channel.
[147] The rendering unit-A 930 applies the rendering information HM_L to the mono
downmix signal to generate a surround signal of the left channel. The rendering unit-B
940 applies the rendering information HM_R to the mono downmix signal to generate
a surround signal of the right channel.
[148] The rendering unit 900 in the drawing does not use a decorrelator. Yet, if the
rendering unit-A 930 and the rendering unit-B 940 performs rendering by using the
rendering information Hmoverall_R and Hmoverall_L defined in Math Figure 12, re-
spectively, it is able to obtain the outputs to which the decorrelator is applied, re-
spectively.
[ 149] Meanwhile, in case of attempting to obtain an output in a stereo signal instead of a
surround signal after completion of the rendering performed on a mono downmix
signal, the following two methods are possible.
[ 150] The first method is that instead of using rendering information for a surround effect,
a value used for a stereo output is used. In this case, it is able to obtain a stereo signal
by modifying only the rendering information in the structure shown in FIG. 3.
[151] The second method is that in a decoding process for generating a multi-channel
signal by using a downmix signal and spatial information, it is able to obtain a stereo
signal by performing the decoding process to only a corresponding step to obtain a
specific channel number.
[152] Referring to FIG. 9, the rendering unit 900 corresponds to a case in which a
decorrelated signal is represented as one, i.e., Math Figure 11. The rendering unit 900
includes a rendering unit-1A 931, a rendering unit-2A 932, a rendering unit-IB 941,
and a rendering unit-2B 942. The rendering unit 900 is similar to the rendering unit for
the stereo downmix signal except that the rendering unit 900 includes the rendering
units 941 and 942 for a decorrelated signal.
[153] In case of the stereo downmix signal, it can be interpreted that one of two channels
is a decorrelated signal. So, without employing additional decorrelators, it is able to
perform a rendering process by using the formerly defined four kinds of rendering in-
formation HL_L, HL_R and the like. In particular, the rendering unit-1A 931 generates
a signal to be delivered to a same channel by applying the rendering information
HM_L to a mono downmix signal. The rendering unit-2A 932 generates a signal to be
delivered to another channel by applying the rendering information HM_R to the mono
downmix signal. The rendering unit-IB 941 generates a signal to be delivered to a
same channel by applying the rendering information HMD_R to a decorrelated signal.
And, the rendering unit-2B 942 generates a signal to be delivered to another channel
by applying the rendering information HMD_L to the decorrelated signal.
[ 154] If a downmix signal is a mono signal, a downmix signal defined as x, source
channel information defined as D, prototype filter information defined as G, a multi-
channel signal defined as p, and a surround signal defined as y can be represented by
matrixes shown in Math Figure 19.
[155] MathFigure 19
[156] In this case, the relation between the matrixes is similar to that of the case that the
downmix signal is the stereo signal. So its details are omitted.
[157] Meanwhile, the source mapping information described with reference to FIG. 4 and
FIG. 5 and the rendering information generated by using the source mapping in-
formation have values differing per frequency band, parameter band, and/or
transmitted timeslot. In this case, if a value of the source mapping information and/or
the rendering information has a considerably big difference between neighbor bands or
between boundary timeslots, distortion may take place in the rendering process. To
prevent the distortion, a smoothing process on a frequency and/or time domain is
needed. Another smoothing method suitable for the rendering is usable as well as the
frequency domain smoothing and/or the time domain smoothing. And, it is able to use
a value resulting from multiplying the source mapping information or the rendering in-
formation by a specific gain.
[158] FIG. 10 and FIG. 11 are block diagrams of a smoothing unit and an expanding unit
according to one embodiment of the present invention.
[159] A smoothing method according to the present invention, as shown in FIG. 10 and
FIG. 11, is applicable to rendering information and/or source mapping information.
Yet, the smoothing method is applicable to other type information. In the following de-
scription, smoothing on a frequency domain is described. Yet, the present invention
includes time domain smoothing as well as the frequency domain smoothing.
[ 160] Referring to FIG. 10 and FIG. 11, the smoothing unit 1042 is capable of performing
smoothing on rendering information and/or source mapping information. A detailed
example of a position of the smoothing occurrence will be described with reference to
FIGs. 18 to 20 later.
[161] The smoothing unit 1042 can be configured with an expanding unit 1043, in which
the rendering information and/or source mapping information can be expanded into a
wider range, for example filter band, than that of a parameter band. In particular, the
source mapping information can be expanded to a frequency resolution (e.g., filter
band) corresponding to filter information to be multiplied by the filter information
(e.g., HRTF filter coefficient). The smoothing according to the present invention is
executed prior to or together with the expansion. The smoothing used together with the
expansion can employ one of the methods shown in FIGs. 12 to 16.
[162] FIG. 12 is a graph to explain a first smoothing method according to one
embodiment of the present invention.
[163] Referring to FIG. 12, a first smoothing method uses a value having the same size as
spatial information in each parameter band. In this case, it is able to achieve a
smoothing effect by using a suitable smoothing function.
[164] FIG. 13 is a graph to explain a second smoothing method according to one
embodiment of the present invention.
[165] Referring to FIG. 13, a second smoothing method is to obtain a smoothing effect by
connecting representative positions of parameter band. The representative position is a
right center of each of the parameter bands, a central position proportional to a log
scale, a bark scale, or the like, a lowest frequency value, or a position previously
determined by a different method.
[166] FIG. 14 is a graph to explain a third smoothing method according to one
embodiment of the present invention.
[167] Referring to FIG. 14, a third smoothing method is to perform smoothing in a form
of a curve or straight line smoothly connecting boundaries of parameters. In this case,
the third smoothing method uses a preset boundary smoothing curve or low pass
filtering by the first order or higher 1TR filter or FIR filter.
[168] FIG. 15 is a graph to explain a fourth smoothing method according to one
embodiment of the present invention.
[169] Referring to FIG. 15, a fourth smoothing method is to achieve a smoothing effect
by adding a signal such as a random noise to a spatial information contour. In this case,
a value differing in channel or band is usable as the random noise. In case of adding a
random noise on a frequency domain, it is able to add only a size value while leaving a
phase value intact. The fourth smoothing method is able to achieve an inter-channel
decorrelation effect as well as a smoothing effect on a frequency domain.
[170] FIG. 16 is a graph to explain a fifth smoothing method according to one
embodiment of the present invention.
[171] Referring to FIG. 16, a fifth smoothing method is to use a combination of the
second to fourth smoothing methods. For instance, after the representative positions of
the respective parameter bands have been connected, the random noise is added and
low path filtering is then applied. In doing so, the sequence can be modified. The fifth
smoothing method minimizes discontinuous points on a frequency domain and an
inter-channel decorrelation effect can be enhanced.
[172] In the first to fifth smoothing methods, a total of powers for spatial information
values (e.g., CLD values) on the respective frequency domains per channel should be
uniform as a constant. For this, after the smoothing method is performed per channel,
power normalization should be performed. For instance, if a downmix signal is a mono
signal, level values of the respective channels should meet the relation of Math Figure
20.
[173] MathFigure 20
D_L(pb) + D_R(pb) + D_C(pb) + D_Ls(pb) + D_Rs(pb) + D_LfeCpb) = C
[174] In this case, 'pb = 0~ total parameter band number 1' and 'C' is an arbitrary constant.
[175] FIG. 17 is a diagram to explain prototype filter information per channel.
[ 176] Referring to FIG. 17, for rendering, a signal having passed through GL_L filter for
a left channel source is sent to a left output, whereas a signal having passed through
GL_R filter is sent to a right output.
[177] Subsequently, a left final output (e.g., Lo) and a right final output (e.g., Ro) are
generated by adding all signals received from the respective channels. In particular, the
rendered left/right channel outputs can be expressed as Math Figure 21.
[178] MathFigure 21
Lo = L * GL_L + C * GC_L + R * GR_L + Ls * GLs_L + Rs * GRs_L
Ro = L * GL_R + C * GC_R + R * GR_R + Ls * GLs + Rs * GRs_R
[179] In the present invention, the rendered left/right channel outputs can be generated by
using the L, R, C, Ls, and Rs generated by decoding the downmix signal into the multi-
channel signal using the spatial information. And, the present invention is able to
generate the rendered left/right channel outputs using the rendering information
without generating the L, R, C, Ls, and Rs, in which the rendering information is
generated by using the spatial information and the filter information.
[180] A process for generating rendering information using spatial information is
explained with reference to FIGs. 18 to 20 as follows.
[181] FIG. 18 is a block diagram for a first method of generating rendering information in
a spatial information converting unit 900 according to one embodiment of the present
invention.
[182] Referring to FIG. 18, as mentioned in the foregoing description, the spatial in-
formation converting unit 900 includes the source mapping unit 1010, the sub-
rendering information generating unit 1020, the integrating unit 1030, the processing
unit 1040, and the domain converting unit 1050. The spatial information converting
unit 900 has the same configuration shown in FIG. 3.
[183] The sub-rendering information generating unit 1020 includes at least one or more
sub-rendering information generating units (1st sub-rendering information generating
unit to Nth sub-rendering information generating unit).
[184] The sub-rendering information generating unit 1020 generates sub-rendering in-
formation by using filter information and source mapping information.
[185] For instance, if a downmix signal is a mono signal, the first sub-rendering in-
formation generating unit is able to generate sub-rendering information corresponding
to a left channel on a multi-channel. And, the sub-rendering information can be
represented as Math Figure 22 using the source mapping information D_L and the
converted filter information GL_L' and GL_R'
[186] MathFigure 22
FL__L = D_L * GL_L'
(mono input filter coefficient to left output channel)
FL_R = D_L * GL_R'
(mono input filter coefficient to right output channel)
[187] In this case, the D_L is a value generated by using the spatial information in the
source mapping unit 1010. Yet, a process for generating the D_L can follow the tree
structure.
[188] The second sub-rendering information generating unit is able to generate sub-
rendering information FR_L and FR_R corresponding to a right channel on the multi-
channel. And, the Nth sub-rendering information generating unit is able to generate sub-
rendering information FRs_L and FRs_R corresponding to a right surround channel on
the multi-channel.
[189] If a downmix signal is a stereo signal, the first sub-rendering information generating
unit is able to generate sub-rendering information corresponding to the left channel on
the multi-channel. And, the sub-rendering information can be represented as Math
Figure 23 by using the source mapping information D_L1 and D_L2.
[190] MathFigure 23
FL_L1 = D_L1 * 6L_L'
(left input filter coefficient to left output channel)
FL_L2 = D_L2 * GL_L'
(right input filter coefficient to left output channel)
FL_R1 = D_L1 * GL_R'
(left input filter coefficient to right output channel)
FL_R2 = D_L2 * GL_R'
(right input filter coefficient to right output channel)
[191] In Math Figure 23, the FL_R1 is explained for example as follows.
[ 192] First of all, in the FL_R1, 'L' indicates a position of the multi-channel, 'R' indicates
an output channel of a surround signal, and T indicates a channel of the downmix
signal. Namely, the FL_R1 indicates the sub-rendering information used in generating
the right output channel of the surround signal from the left channel of the downmix
signal.
[193] Secondly, the D_L1 and the D_L2 are values generated by using the spatial in-
formation in the source mapping unit 1010.
[194] If a downmix signal is a stereo signal, it is able to generate a plurality of sub-
rendering informations from at least one sub-rendering information generating unit in
the same manner of the case that the downmix signal is the mono signal. The types of
the sub-rendering informations generated by a plurality of the sub-rendering in-
formation generating units are exemplary, which does not put limitation on the present
invention.
[195] The sub-rendering information generated by the sub-rendering information
generating unit 1020 is transferred to the rendering unit 900 via the integrating unit
1030, the processing unit 1040, and the domain converting unit 1050.
[196] The integrating unit 1030 integrates the sub-rendering informations generated per
channel into rendering information (e.g., HL_L, HL_R, HR_L, HR_R) for a rendering
process. An integrating process in the integrating unit 1030 is explained for a case of a
mono signal and a case of a stereo signal as follows.
[197] First of all, if a downmix signal is a mono signal, rendering information can be
expressed as Math Figure 24.
[198] MathFigure 24
HM_L = FL_L + FR_L + FC_L + FLs_L + FRs_L + FLFE_L
HO = FL_R + FR_R + FC_R + FLs_R + FRs.R + FLFE_R
[199] Secondly, if a downmix signal is a stereo signal, rendering information can be
expressed as Math Figure 25.
[200] MathFigure 25
HL_L - FL_L1 + FR_L1 4 FC_L1 + FLs_Ll + FRs_Ll + FLFE_L1
HR_L = FL_L2 + FR_L2 + FC_L2 + FLs_L2 + FRs_L2 4 FLFE_L2
HL_R = FL_R1 + FR_R1 + FC_R1 + FLs_Rl + FRs_Rl + FLFE_R1
HR_R - FL_R2 + FR_R2 + FC_R2 + FLs_R2 + FRs_R2 + FLFE_R2
[201] Subsequently, the processing unit 1040 includes an interpolating unit 1041 and/or a
smoothing unit 1042 and performs interpolation and/or smoothing for the rendering in-
formation. The interpolation and/or smoothing can be executed on a time domain, a
frequency domain, or a QMF domain. In the specification, the time domain is taken as
an example, which does not put limitation on the present invention.
[202] The interpolation is performed to obtain rendering information non-existing
between the rendering informations if the transmitted rendering information has a wide
interval on the time domain. For instance, assuming that rendering informations exist
in an nth timeslot and an (n+k)th timeslot (k>l), respectively, it is able to perform linear
interpolation on a not-transmitted timeslot by using the generated rendering in-
formations (e.g., HL_L, HR_L, HL_R, HR_R).
[203] The rendering information generated from the interpolation is explained with
reference to a case that a downmix signal is a mono signal and a case that the downmix
signal is a stereo signal.
[204] If the downmix signal is the mono signal, the interpolated rendering information
can be expressed as Math Figure 26.
[205] MathFigure 26
HM_L(n+j) = HM_L(n) *(l-a) + HM_L(n4k) * a
HM_R(n+j) = HM_R(n) *(l-a) + HM_R(n+k) * a
[206] If the downmix signal is the stereo signal, the interpolated rendering information
can be expressed as Math Figure 27.
[207] MathFigure 27
HL_L(n+j) = HL_L(n) *(l-a) + HL_L(n+k) * a
HR_L(n+j) = HR_L(n) *(l-a) + HR_L(n+k) * a
HL_R(n+j) = HL_R(n) *(l-a) + HL_R(n+k) * a
HR_R(n+j) = HR_R(n) *(l-a) + HR_R(n+k) * a
[208] In this case, it is 0
responding to '0
[209] MathFigure 28
a = j/k
[210] If so, it is able to obtain a value corresponding to the not-transmitted timeslot on a
straight line connecting the values in the two timeslots according to Math Figure 27
and Math Figure 28. Details of the interpolation will be explained with reference to
FIG. 22 and FIG. 23 later.
[211] In case that a filter coefficient value abruptly varies between two neighboring
timeslots on a time domain, the smoothing unit 1042 executes smoothing to prevent a
problem of distortion due to an occurrence of a discontinuous point. The smoothing on
the time domain can be carried out using the smoothing method described with
reference to FIGs. 12 to 16. The smoothing can be performed together with expansion.
And, the smoothing may differ according to its applied position. If a downmix signal is
a mono signal, the time domain smoothing can be represented as Math Figure 29.
[212] MathFigure 29
HM_L(n)' = HM_L(n)*b + HM_L(n-l)'*(1-b)
HM_R(n)' = HM_R(n)*b + HM_R(n-l) '*(1-b)
[213] Namely, the smoothing can be executed by the 1-pol IIR filter type performed in a
manner of multiplying the rendering information HM_L(n-l) or HM_R(n-l) smoothed
in a previous timeslot n-1 by (1-b), multiplying the rendering information HM_L(n) or
HM)R(n) generated in a current timeslot n by b, and adding the two multiplications
together. In this case, 'b' is a constant for 0
becomes greater. If 'b' gets bigger, a smoothing effect becomes smaller. And, the rest
of the filters can be applied in the same manner.
[214] The interpolation and the smoothing can be represented as one expression shown in
Math Figure 30 by using Math Figure 29 for the time domain smoothing.
[215] MathFigure 30
HM_L(n+j)'=(HM_L(n)*(1 -a)+HM_L(n+k)*a)*b+HM_L(n+j-1 )'*(1 -b)
HM_R(n+j)'=(HM_R(n)*(1-a)+HM_R(n+k)*a)*b+HM_R(n+j-1)'*(1-b)
[216] If the interpolation is performed by the interpolating unit 1041 and/or if the
smoothing is performed by the smoothing unit 1042, rendering information having an
energy value different from that of prototype rendering information may be obtained.
To prevent this problem, energy normalization may be executed in addition.
[217] Finally, the domain converting unit 1050 performs domain conversion on the
rendering information for a domain for executing the rendering. If the domain for
executing the rendering is identical to the domain of rendering information, the domain
conversion may not be executed. Thereafter, the domain-converted rendering in-
formation is transferred to the rendering unit 900.
[218] FIG. 19 is a block diagram for a second method of generating rendering information
in a spatial information converting unit according to one embodiment of the present
invention.
[219] The second method is similar to the first method in that a spatial information
converting unit 1000 includes a source mapping unit 1010, a sub-rendering information
generating unit 1020, an integrating unit 1030, a processing unit 1040, and a domain
converting unit 1050 and in that the sub-rendering information generating unit 1020
includes at least one sub-rendering information generating unit.
[220] Referring to FIG. 19, the second method of generating the rendering information
differs from the first method in a position of the processing unit 1040. So, interpolation
and/or smoothing can be performed per channel on sub-rendering informations (e.g.,
FLJL and FL_R in case of mono signal or FL_L1, FL_L2 ,FL_R1, FL_R2 in case of
stereo signal) generated per channel in the sub-rendering information generating unit
1020.
[221] Subsequently, the integrating unit 1030 integrates the interpolated and/or smoothed
sub-rendering informations into rendering information.
[222] The generated rendering information is transferred to the rendering unit 900 via the
domain converting unit 1050.
[223] FIG. 20 is a block diagram for a third method of generating rendering filter in-
formation in a spatial information converting unit according to one embodiment of the
present invention.
[224] The third method is similar to the first or second method in that a spatial in-
formation converting unit 1000 includes a source mapping unit 1010, a sub-rendering
information generating unit 1020, an integrating unit 1030, a processing unit 1040, and
a domain converting unit 1050 and in that the sub-rendering information generating
unit 1020 includes at least one sub-rendering information generating unit.
[225] Referring to FIG. 20, the third method of generating the rendering information
differs from the first or second method in that the processing unit 1040 is located next
to the source mapping unit 1010. So, interpolation and/or smoothing can be performed
per channel on source mapping information generated by using spatial information in
the source mapping unit 1010.
[226] Subsequently, the sub-rendering information generating unit 1020 generates sub-
rendering information by using the interpolated and/or smoothed source mapping in-
formation and filter information.
[227] The sub-rendering information is integrated into rendering information in the in-
tegrating unit 1030. And, the generated rendering information is transferred to the
rendering unit 900 via the domain converting unit 1050.
[228] FIG. 21 is a diagram to explain a method of generating a surround signal in a
rendering unit according to one embodiment of the present invention. FIG. 21 shows a
rendering process executed on a DFT domain. Yet, the rendering process can be im-
plemented on a different domain in a similar manner as well. FIG. 21 shows a case that
an input signal is a mono downmix signal. Yet, FIG. 21 is applicable to other input
channels including a stereo downmix signal and the like in the same manner.
[229] Referring to FIG. 21, a mono downmix signal on a time domain preferentially
executes windowing having an overlap interval OL in the domain converting unit. Fig.
21 shows a case that 50% overlap is used. Yet, the present invention includes cases of
using other overlaps.
[230] A window function for executing the windowing can employ a function having a
good frequency selectivity on a DFT domain by being seamlessly connected without
discontinuity on a time domain. For instance, a sine square window function can be
used as the window function.
[231] Subsequently, zero padding ZL of a tab length [precisely, (tab length) -1 ] of a
rendering filter using rendering information converted in the domain converting unit is
performed on a mono downmix signal having a length OL*2 obtained from the
windowing. A domain conversion is then performed into a DFT domain. FIG. 20
shows that a block-k downmix signal is domain-converted into a DFT domain.
[232] The domain-converted downmix signal is rendered by a rendering filter that uses
rendering information. The rendering process can be represented as a product of a
downmix signal and rendering information. The rendered downmix signal undergoes
IDFT (Inverse Discrete Fourier Transform) in the inverse domain converting unit and
is then overlapped with the downmix signal (block k-1 in FIG. 20) previously executed
with a delay of a length OL to generate a surround signal.
[233] Interpolation can be performed on each block undergoing the rendering process.
The interpolating method is explained as follows.
[234] FIG. 22 is a diagram for a first interpolating method according to one embodiment
of the present invention. Interpolation according to the present invention can be
executed on various positions. For instance, the interpolation can be executed on
various positions in the spatial information converting unit shown in FIGs. 18 to 20 or
can be executed in the rendering unit. Spatial information, source mapping in-
formation, filter information and the like can be used as the values to be interpolated.
In the specification, the spatial information is exemplarily used for description. Yet,
the present invention is not limited to the spatial information. The interpolation is
executed after or together with expansion to a wider band.
[235] Referring to FIG. 22, spatial information transferred from an encoding apparatus c
an be transferred from a random position instead of being transmitted each timeslot.
One spatial frame is able to carry a plurality of spatial information sets (e.g., parameter
sets n and n+1 in FIG. 22). In case of a low bit rate, one spatial frame is able to carry a
single new spatial information set. So, interpolation is carried out for a not-transmitted
timeslot using values of a neighboring transmitted spatial information set. An interval
between windows for executing rendering does not always match a timeslot. So, an in-
terpolated value at a center of the rendering windows (K-l, K, K+l, K+2, etc.), as
shown in FIG. 22, is found to use. Although FIG. 22 shows that linear interpolation is
carried out between timeslots where a spatial information set exists, the present
invention is not limited to the interpolating method. For instance, interpolation is not
carried out on a timeslot where a spatial information set does not exist. Instead, a
previous or preset value can be used.
[236] FIG. 23 is a diagram for a second interpolating method according to one
embodiment of the present invention.
[237] Referring to FIG. 23, a second interpolating method according to one embodiment
of the present invention has a structure that an interval using a previous value, an
interval using a preset default value and the like are combined. For instance, in-
terpolation can be performed by using at least one of a method of maintaining a
previous value, a method of using a preset default value, and a method of executing
linear interpolation in an interval of one spatial frame. In case that at least two new
spatial information sets exist in one window, distortion may take place. In the
following description, block switching for preventing the distortion is explained.
[238] FIG. 24 is a diagram for a block switching method according to one embodiment of
the present invention.
[239] Referring to (a) shown in FIG. 24, since a window length is greater than a timeslot
length, at least two spatial information sets (e.g., parameter sets n and n+1 in FIG. 24)
can exist in one window interval. In this case, each of the spatial information sets
should be applied to a different timeslot. Yet, if one value resulting from interpolating
the at least two spatial information sets is applied, distortion may take place. Namely,
distortion attributed to time resolution shortage according to a window length can take
place.
[240] To solve this problem, a switching method of varying a window size to fit
resolution of a timeslot can be used. For instance, a window size, as shown in (b) of
FIG. 24, can be switched to a shorter-sized window for an interval requesting a high
resolution. In this case, at a beginning and an ending portion of switched windows,
connecting windows is used to prevent seams from occurring on a time domain of the
switched windows.
[241] The window length can be decided by using spatial information in a decoding
apparatus instead of being transferred as separate additional information. For instance,
a window length can be determined by using an interval of a timeslot for updating
spatial information. Namely, if the interval for updating the spatial information is
narrow, a window, function of short length is used. If the interval for updating the
spatial information is wide, a window function of long length is used. In this case, by
using a variable length window in rendering, it is advantageous not to use bits for
sending window length information separately. Two types of window length are shown
in (b) of FIG. 24. Yet, windows having various lengths can be used according to
transmission frequency and relations of spatial information. The decided window
length information is applicable to various steps for generating a surround signal,
which is explained in the following description.
[242] FIG. 25 is a block diagram for a position to which a window length decided by a
window length deciding unit is applied according to one embodiment of the present
invention.
[243] Referring to FIG. 25, a window length deciding unit 1400 is able to decide a
window length by using spatial information. Information for the decided window
length is applicable to a source mapping unit 1010, an integrating unit 1030, a
processing unit 1040, domain converting units 1050 and 1100, and a inverse domain
converting unit 1300. Fig. 25 shows a case that a stereo downmix signal is used. Yet,
the present invention is not limited to the stereo downmix signal only. As mentioned in
the foregoing description, even if a window length is shortened, a length of zero
padding decided according to a filter tab number is not adjustable. So, a solution for
the problem is explained in the following description.
[244] FIG. 26 is a diagram for filters having various lengths used in processing an audio
signal according to one embodiment of the present invention. As mentioned in the
foregoing description, if a length of zero padding decided according to a filter tab
number is not adjusted, an overlapping amounting to a corresponding length sub-
stantially occurs to bring about time resolution shortage. A solution for the problem is
to reduce the length of the zero padding by restricting a length of a filter tab. A method
of reducing the length of the zero padding can be achieved by truncating a rear portion
of a response (e.g., a diffusing interval corresponding to reverberation). In this case, a
rendering process may be less accurate than a case of not truncating the rear portion of
the filter response. Yet, filter coefficient values on a time domain are very small to
mainly affect reverberation. So, a sound quality is not considerably affected by the
truncating.
[245] Referring to FIG. 26, four kinds of filters are usable. The four kinds of the filters are
usable on a DFT domain, which does not put limitation on the present invention.
[246] A filter-N indicates a filter having a long filter length FL and a length 2*OL of a
long zero padding of which filter tab number is not restricted. A filter-N2 indicates a
filter having a zero padding length 2*OL shorter than that of the filter-Nl by restricting
a tab number of filter with the same filter length FL. A filter-N3 indicates a filter
having a long zero padding length 2*OL by not restricting a tab number of filter with a
filter length FL shorter than that of the filter-Nl. And, a filter-N4 indicates a filter
having a window length FL shorter than that of the filter-Nl with a short zero padding
length 2*OL by restricting a tab number of filter.
[247] As mentioned in the foregoing description, it is able to solve the problem of time
resolution using the above exemplary four kinds of the filters. And, for the rear portion
of the filter response, a different filter coefficient is usable for each domain.
[248] FIG. 27 is a diagram for a method of processing an audio signal dividedly by using
a plurality of subfilters according to one embodiment of the present invention, one
filter may be divided into subfilters having filter coefficients differing from each other.
After processing the audio signal by using the subfilters, a method of adding results of
the processing can be used. In case applying spatial information to a rear portion of a
filter response having small energy, i.e., in case of performing rendering by using a
filter with a long filter tab, the method provides function for processing dividedly the
audio signal by a predetermined length unit. For instance, since the rear portion of the
filter response is not considerably varied per HRTF corresponding to each channel, it is
able to perform the rendering by extracting a coefficient common to a plurality of
windows. In the present specification, a case of execution on a DFT domain is
described. Yet, the present invention is not limited to the DFT domain.
[249] Referring to FIG. 27, after one filter FL has been divided into a plurality of sub-
areas, a plurality of the sub-areas can be processed by a plurality of subfilters (filter-A
and filter-B) having filter coefficients differing from each other.
[250] Subsequently, an output processed by the filter-A and an output processed by the
filter-B are combined together. For instance, IDFT (Inverse Discrete Fourier
Transform) is performed on each of the output processed by the filter-A and the output
processed by the filter-B to generate a time domain signal. And, the generated signals
are added together. In this case, a position, to which the output processed by the filter-
B is added, is time-delayed by FL more than a position of the output processed by the
filter-A. In this way, the signal processed by a plurality of the subfilters brings the
same effect of the case that the signal is processed by a single filter.
[251 ] And, the present invention includes a method of rendering the output processed by
the filter-B to a downmix signal directly. In this case, it is able to render the output to
the downmix signal by using coefficients extracting from spatial information, the
spatial information in part or without using the spatial information.
[252] The method is characterized in that a filter having a long tab number can be applied
dividedly and that a rear portion of the filter having small energy is applicable without
conversion using spatial information. In this case, if conversion using spatial in-
formation is not applied, a different filter is not applied to each processed window. So,
it is unnecessary to apply the same scheme as the block switching. FIG. 26 shows that
the filter is divided into two areas. Yet, the present invention is able to divide the filter
into a plurality of areas.
[253] FIG. 28 is a block diagram for a method of rendering partition rendering in-
formation generated by a plurality of subfilters to a mono downmix signal according to
one embodiment of the present invention. FIG. 28 relates to one rendering coefficient.
The method can be executed per rendering coefficient.
[254] Referring to FIG. 28, the filter-A information of FIG. 27 corresponds to first
partition rendering information HM_L_A and the filter-B information of FIG. 27
corresponds to second partition rendering information HM_L_B. FIG. 28 shows an
embodiment of partition into two subfilters. Yet, the present invention is not limited to
the two subfilters. The two subfilters can be obtained via a splitting unit 1500 using the
rendering information HM_L generated in the spatial information generating unit
1000. Alternatively, the two subfilters can be obtained using prototype HRTF in-
formation or information decided according to a user's selection. The information
decided according to a user's selection may include spatial information selected
according to a user's taste for example. In this case, HM_L_A is the rendering in-
formation based on the received spatial information, and, HM_L_B may be the
rendering information for providing a 3-dimensional effect commonly applied to
signals.
[255] As mentioned in the foregoing description, the processing with a plurality of the
subfilters is applicable to a time domain and a QMF domain as well as the DFT
domain. In particular, the coefficient values split by the filter-A and the filter-B are
applied to the downmix signal by time or QMF domain rendering and are then added
to generate a final signal.
[256] The rendering unit 900 includes a first partition rendering unit 950 and a second
partition rendering unit 960. The first partition rendering unit 950 performs a rendering
process using HM_L_A, whereas the second partition rendering unit 960 performs a
rendering process using HM_L_B.
[257] If the filter-A and the filter-B, as shown in FIG. 27, are splits of a same filter
according to time, it is able to consider a proper delay to correspond to the time
interval. FIG. 28 shows an example of a mono downmix signal. In case of using mono
downmix signal and decorrelator, a portion corresponding to the filter-B is applied not
to the decorrelator but to the mono downmix signal directly.
[258] FIG. 29 is a block diagram for a method of rendering partition rendering in-
formation generated using a plurality of subfilters to a stereo downmix signal
according to one embodiment of the present invention.
[259] A partition rendering process shown in FIG. 29 is similar to that of FIG. 28 in that
two subfilters are obtained in a splitter 1500 by using rendering information generated
by the spatial information converting unit 1000, prototype HRTF filter information or
user decision information. The difference from FIG. 28 lies in that a partition rendering
process corresponding to the filter-B is commonly applied to L/R signals.
[260] In particular, the splitter 1500 generates first partition rendering information cor-
responding to filter-A information, second partition rendering information, and third
partition rendering information corresponding to filter-B information. In this case, the
third partition rendering information can be generated by using filter information or
spatial information commonly applicable to the L/R signals.
[261 ] Referring to FIG. 29, a rendering unit 900 includes a first partition rendering unit
970, a second partition rendering unit 980, and a third partition rendering unit 990.
[262] The third partition rendering information generates is applied to a sum signal of the
L/R signals in the third partition rendering unit 990 to generate one output signal. The
output signal is added to the L/R output signals, which are independently rendered by a
filter-A1 and a filter-A2 in the first and second partition rendering units 970 and 980,
respectively, to generate surround signals. In this case, the output signal of the third
partition rendering unit 990 can be added after an appropriate delay. In FIG. 29, an
expression of cross rendering information applied to another channel from L/R inputs
is omitted for convenience of explanation.
[263] FIG. 30 is a block diagram for a first domain converting method of a downmix
signal according to one embodiment of the present invention. The rendering process
executed on the DFT domain has been described so far. As mentioned in the foregoing
description, the rendering process is executable on other domains as well as the DFT
domain. Yet, FIG. 30 shows the rendering process executed on the DFT domain. A
domain converting unit 1100 includes a QMF filter and a DFT filter. An inverse
domain converting unit 1300 includes an IDFT filter and an IQMF filter. FIG. 30
relates to a mono downmix signal, which does not put limitation on the present
invention.
[264] Referring to Fig. 30, a time domain downmix signal of p samples passes through a
QMF filter to generate P sub-band samples. W samples are recollected per band. After
windowing is performed on the recollected samples, zero padding is performed. M-
point DFT (FFT) is then executed. In this case, the DFT enables a processing by the
aforesaid type windowing. A value connecting the M/2 frequency domain values per
band obtained by the M-point DFT to P bands can be regarded as an approximate value
of a frequency spectrum obtained by M/2*P-point DFT. So, a filter coefficient
represented on a M/2*P-point DFT domain is multiplied by the frequency spectrum to
bring the same effect of the rendering process on the DFT domain.
[265] In this case, the signal having passed through the QMF filter has leakage, e.g.,
aliasing between neighboring bands. In particular, a value corresponding to a neighbor
band smears in a current band and a portion of a value existing in the current band is
shifted to the neighbor band. In this case, if QMF integration is executed, an original
signal can be recovered due to QMF characteristics. Yet, if a filtering process is
performed on the signal of the corresponding band as the case in the present invention,
the signal is distorted by the leakage. To minimize this problem, a process for
recovering an original signal can be added in a manner of having a signal pass through
a leakage minimizing butterfly B prior to performing DFT per band after QMF in the
domain converting unit 100 and performing a reversing process V after IDFT in the
inverse domain converting unit 1300.
[266] Meanwhile, to match the generating process of the rendering information generated
in the spatial information converting unit 1000 with the generating process of the
downmix signal, DFT can be performed on a QMF pass signal for prototype filter in-
formation instead of executing M/2*P-point DFT in the beginning. In this case, delay
and data spreading due to QMF filter may exist.
[267] FIG. 31 is a block diagram for a second domain converting method of a downmix
signal according to one embodiment of the present invention. FIG. 31 shows a
rendering process performed on a QMF domain.
[268] Referring to FIG. 31, a domain converting unit 1100 includes a QMF domain
converting unit and an inverse domain converting unit 1300 includes an IQMF domain
converting unit. A configuration shown in FIG. 31 is equal to that of the case of using
DFT only except that the domain converting unit is a QMF filter. In the following de-
scription, the QMF is referred to as including a QMF and a hybrid QMF having the
same bandwidth. The difference from the case of using DFT only lies in that the
generation of the rendering information is performed on the QMF domain and that the
rendering process is represented as a convolution instead of the product on the DFT
domain, since the rendering process performed by a renderer-M 3012 is executed on
the QMF domain.
[269] Assuming that the QMF filter is provided with B bands, a filter coefficient can be
represented as a set of filter coefficients having different features (coefficients) for the
B bands. Occasionally, if a filter tab number becomes a first order (i.e., multiplied by a
constant), a rendering process on a DFT domain having B frequency spectrums and an
operational process are matched. Math Figure 31 represents a rendering process
executed in one QMF band (b) for one path for performing the rendering process using
rendering information HM_L.
[270] MathFigure 31
[271] In this case, k indicates a time order in QMF band, i.e., a timeslot unit. The
rendering process executed on the QMF domain is advantageous in that, if spatial in-
formation transmitted is a value applicable to the QMF domain, application of cor-
responding data is most facilitated and that distortion in the course of application can
be minimized. Yet, in case of QMF domain conversion in the prototype filter in-
formation (e.g., prototype filter coefficient) converting process, a considerable op-
erational quantity is required for a process of applying the converted value. In this
case, the operational quantity can be minimized by the method of parameterizing the
HRTF coefficient in the filter information converting process.
Industrial Applicability
[272] Accordingly, the signal processing method and apparatus of the present invention
uses spatial information provided by an encoder to generate surround signals by using
HRTF filter information or filter information according to a user in a decoding
apparatus in capable of generating multi-channels. And, the present invention is
usefully applicable to various kinds of decoders capable of reproducing stereo signals
only.
[273] While the present invention has been described and illustrated herein with reference
to the preferred embodiments thereof, it will be apparent to those skilled in the art that
various modifications and variations can be made therein without departing from the
spirit and scope of the invention. Thus, it is intended that the present invention covers
the modifications and variations of this invention that come within the scope of the
appended claims and their equivalents.
Claims
[I] A method of processing a signal, the method comprising:
generating source mapping information corresponding to each source of multi-
sources by using spatial information indicating features between the multi-
sources;
generating sub-rendering information by applying filter information having a
surround effect to the source mapping information per the source;
generating rendering information for generating a surround signal by integrating
at least one of the sub-rendering information; and
generating the surround signal by applying the rendering information to a
downmix signal generated by downmixing the multi-sources.
[2] The method of claim 1, wherein the spatial information includes at least one of a
source level difference and an inter-source correlation.
[3] The method of claim 2, wherein the source mapping information includes
decorrelating information.
[4] The method of claim 3, wherein the decorrelating information corresponds to
each source of the multi-sources.
[5] The method of claim 4, wherein the decorrelating information is configured in a
manner that the decorrelating information is symmetric for left source and right
source of the multi-sources, and that the decorrelating information is symmetric
for left surround source and right surround source of the multi-sources.
[6] The method of claim 3, wherein the decorrelating information has an all-pass
feature.
[7] The method of claim 6, wherein the decorrelating information has a serial or
overlapped structural feature.
[8] The method of claim 2, wherein the source mapping information is generated by
using features of a decorrelator applied to a frequency domain or a domain
different from the frequency domain.
[9] The method of claim 1, wherein the sub-rendering information includes in-
formation generated by applying the filter information to at least two of the
source mapping information corresponding to each source of the multi-sources.
[ 10] The method of claim 1, wherein the filter information includes HRTF filter in-
formation or a value decided according to a user's selection.
[II] The method of claim 10, wherein the filter information is domain-converted into
a domain for generating the surround signal.
[12] The method of claim 11, wherein the filter information is generated by
converting the HRTF filter information into a parameter.
[13] An apparatus for processing a signal, the apparatus comprising:
a source mapping unit generating source mapping information corresponding to
each source of multi-sources by using spatial information indicating features
between the multi-sources;
a sub-rendering information generating unit generating sub-rendering in-
formation by applying filter information having a surround effect to the source
mapping information per the source;
an integrating unit generating rendering information for generating a surround
signal by integrating the at least one of the sub-rendering information; and
a rendering unit generating the surround signal by applying the rendering in-
formation to a downmix signal generated by downmixing the multi-sources.
[14] The apparatus of claim 13, wherein the spatial information includes at least one
of a source level difference and an inter-source correlation.
[15] The apparatus of claim 13, wherein the sub-rendering information generating
unit generates the sub-rendering information by applying the filter information to
at least two of the source mapping information corresponding to each source of
the multi-sources.
[16] The apparatus of claim 13, further comprising a filter information converting unit
converting a domain of the filter information to a domain for generating the
surround signal, in which the filter information includes HRTF filter information
or a value decided according to a user's selection.
An apparatus for processing a media signal and method thereof are disclosed, by which the media signal can be converted to a surround signal by using spatial information of the media signal. The present invention provides a method of processing a signal, the method comprising of generating source mapping information corresponding to each source of
multi-sources by using spatial information
indicating features between the multi-sources;
generating sub-rendering information by
applying filter information giving a surround
effect to the source mapping information per
the source; generating rendering information for
generating a surround signal by integrating the
at least one of the sub-rendering information;
and generating the surround signal by applying
the rendering information to a downmix signal
generated by downmixing the multi- sources. |