JP2006337767A - Device and method for parametric multichannel decoding with low operation amount - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that an operation amount and memory capacity increase as data are processed in a complex number region since conventional parametric stereo coding based upon MPEG-4 Audio Extension 2 uses a complex value filter bank. <P>SOLUTION: A parametric multichannel decoding device according to the present invention employs real-number values as at least one or more of filter coefficients that an analyzing filter, an all-pass filter, and a synthesizing filter use to greatly reduce the operation amount and memory capacity. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus and method for executing parametric multi-channel decoding processing compliant with MPEG-4 Audio Extension 2 with a low amount of computation.

  Converting a multi-channel signal into a signal having a smaller number of channels is generally called downmix. Parametric stereo coding processing conforming to MPEG-4 Audio Extension 2 standardized by MPEG (Moving Picture Expert Group) is not limited to this, but in particular, it is a broad transmission that transmits and receives a stereo signal by downmixing it to a monaural signal. This is useful for low bit rate applications such as casting (see Non-Patent Document 1, for example).

  The principle of downmix by parametric stereo coding processing based on MPEG-4 Audio Extension 2 is based on binaural cue coding. A binaural cue is a signal extracted when a stereo signal is downmixed to a monaural signal, and is an IID (inter-channel level / intensity difference: inter-channel level / intensity difference) Difference) queue information, IPD (interchannel phase / delay difference) queue information, and ICC (interchannel coherence / correlation) queue information.

  The IID queue information is a signal indicating the ratio of the magnitudes of the two signals. The IPD queue information is a signal indicating a phase difference between two signals. The ICC queue information is a signal indicating the similarity between two signals (in this case, the left channel and the right channel). In general, the IID cue information representing the level / intensity ratio of two signals and the IPD cue information representing the phase / delay difference between the two signals control the sound balance and localization direction, and interfere with the two signals. The ICC cue information representing the sex / correlation is useful as a parameter for controlling the width and spread of the sound image. These cue information constitutes spatial parameters that support the listener psychologically constructing an auditory scene.

  FIG. 6 is a diagram for explaining the outline of the principle of parametric stereo coding. As shown in FIG. 6, the conventional parametric stereo coding is an encoding device that performs an encoding process including a downmix unit 700, a binaural cue extraction unit 702, an audio signal encoding unit 704, and a multiplexing unit 706, It is executed by a decoding device that performs a decoding process including a demultiplexing unit 708, an audio signal decoding unit 710, and a stereo decoding unit 712 (see, for example, Patent Document 2).

  First, an encoding process executed by the encoding device will be described.

  In the encoding device, the signal is processed in units of frames. The downmix unit 700 generates a monaural signal M by inputting a stereo signal composed of a left channel signal L and a right channel signal R from an input unit (not shown). The monaural signal M generated here is obtained by mixing the left channel signal L and the right channel signal R at a predetermined ratio.

  Next, the binaural cue extraction unit 702 generates the binaural cue BCInfo by inputting the left channel signal L and the right channel signal R from the input unit and the monaural signal M from the downmix unit 700.

  The audio signal encoding unit 704 encodes the monaural signal M to generate a compressed bit stream. The audio signal encoding unit 704 is configured by, for example, an audio encoder conforming to MP3 (MPEG 1 Audio Layer 3) or AAC (Advanced Audio Coding).

  The multiplexing unit 706 outputs the monaural signal M encoded and compressed by the audio signal encoding unit 704 and the binaural cue BCInfo extracted by the binaural cue extraction unit 702 and quantized by a quantization unit (not shown). The bit stream is generated by multiplexing.

  Next, a decoding process executed by the decoding device will be described.

  In the decoding device, when the bit stream generated by the encoding device is input to the decoding device, the demultiplexing unit 708 constituting the decoding device is quantized with the monaural signal M ′ encoded and compressed from the bit stream. Separate binaural cue BCInfo.

  The audio signal decoding unit 710 receives the encoded and compressed monaural signal M ′, decodes it, and restores the monaural signal M ′. The audio signal decoding unit 710 is configured by an audio decoder conforming to a method corresponding to the audio signal encoding unit 704.

  The stereo decoding unit 712 combines the binaural cue BCInfo separated by the demultiplexing unit 708 and dequantized by the dequantization unit (not shown) and the monaural signal M ′ restored by the audio signal decoding unit 710. A stereo signal composed of the left channel signal L ′ and the right channel signal R ′ is restored.

  With reference to FIG. 7, the configuration of stereo decoding section 712 will be described in more detail.

  The stereo decoding unit 712 includes a complex analysis filter unit 800, a complex reverberation signal generation unit 806, an interpolation unit 804, a signal distribution unit 808, and a complex synthesis filter unit 812.

The complex analysis filter unit 800 converts the restored monaural signal M ′, which is a time signal, into a subband signal composed of 64 subband (frequency band) signals each having 32 samples, and further, a predetermined low frequency signal. The subband signal of the region is further converted into a plurality of subsubband signals. Complex valued filter coefficients complex analysis filter unit 800 is used is Mri (k, n) and _{G qi} ^p.

  The interpolation unit 804 generates a processing coefficient by performing linear interpolation before actually applying the binaural queue BCInfo.

Complex reverberation signal generator 806, the mono signal based on the ICC cue information contained in both ears queue BCInfo interpolated by monaural signal _S k (z) and the interpolation unit 804 that has been processed by the complex analysis filter unit 800 _S k ( A reverberation signal D _k (z) uncorrelated with z) is generated. The complex reverberation signal generation unit 806 has an all-pass filter, and a complex-value filter coefficient H _ki (z) characterized by having the following non-integer delay is applied to the all-pass filter.

The signal distribution unit 808 performs complex reverberation on the monaural signal S _k (z) processed by the complex analysis filter unit 800 based on the ICC queue information and the IID queue information included in the binaural queue BCInfo interpolated by the interpolation unit 804. The reverberation signal D _k (z) generated by the signal generation unit 806 is added to generate the left channel signal L ′ and the right channel signal R ′. Here, the ratio of adding the reverberation signal D _k (z) to the monaural signal S _k (z) processed by the complex analysis filter unit 800 is determined based on the ICC queue information, and the reverberation signal D _k (z) is monaural. A ratio to be distributed to the left channel signal L ′ and the right channel signal R ′ in addition to the signal S _k (z) is determined based on the IID queue information.

  The complex synthesis filter unit 812 includes a filter bank, and converts the left channel signal L ′ and the right channel signal R ′ from a time frequency signal to a time signal.

  The filter coefficients Nri (k, n) used by the complex synthesis filter unit 812 are as follows.

ISO / IEC 14496-3: 2001 / FDAM2, "Parametric Coding for high Quality Audio" WO03 / 007656A1, "Efficient and Scalable Parametric Stereo Coding for Low Bitrate Application"

  However, since the stereo decoding unit used in the conventional parametric stereo coding uses a plurality of filters, and each uses complex filter coefficients, there is a problem that the amount of calculation, the processing output, and the memory capacity increase. there were.

  The present invention reduces the requirements for memory, computation amount, and processing output, and downmixed monaural signals and the like that can be used in devices with limited battery capacity such as portable terminals and low bit rate applications. An object of the present invention is to provide a parametric multi-channel decoding device that decodes a multi-channel signal such as a stereo signal from a signal.

  The parametric multi-channel decoding device of the present invention includes an analysis filter unit that converts a downmix signal into a plurality of subband signals using an analysis filter coefficient, and an interpolation that generates a processing coefficient by performing linear interpolation on binaural cue information And an all-pass filter that generates an uncorrelated reverberation signal with the sub-band signal using an all-pass filter coefficient based on the sub-band signal converted by the analysis filter means and the processing coefficient generated by the interpolation means A reverberation signal generating means, a signal distribution means for generating a multichannel signal by mixing the subband signal and the reverberation signal generated by the reverberation signal generating means based on the processing coefficient, and a synthesis filter coefficient Convert the multi-channel signal from subband signal to time signal using In the parametric multi-channel decoding device, the parametric multi-channel decoding device further includes the analysis filter coefficient used by the analysis filter unit and the all-pass filter unit forming the reverberation signal generation unit. One or more of the all-pass filter coefficient and the synthesis filter coefficient used by the synthesis filter means are used as real values.

  Here, converting a multi-channel signal into a signal with a reduced number of channels is called “downmix”, and a signal with a reduced number of channels converted by downmixing from a multichannel signal with a predetermined number of channels. It shall be called “downmix signal”.

  With this configuration, the parametric multi-channel decoding device according to the present invention can perform filtering by a real number calculation with a low load by reducing a complex number calculation with a large load when decoding a multi-channel signal. In a device with limited battery capacity or a low bit rate application such as broadcasting for down-mixing a stereo signal into a monaural signal, it is possible to reduce the requirements for memory, calculation amount, and processing output.

  Further, the parametric multi-channel decoding device of the present invention uses the following real values, that is, the filter coefficients used by the analysis filter means:

および

and

とする構成を有している。

The configuration is as follows.

  With this configuration, the parametric multi-channel decoding device according to the present invention can perform filter processing by real number calculation with a light load on the analysis filter means, so that requirements for memory, calculation amount, and processing output can be reduced.

  Further, the parametric multi-channel decoding device of the present invention uses an all-pass filter coefficient used by the all-pass filter unit forming the reverberation signal generating means as the following real value, that is,

とする構成を有している。

The configuration is as follows.

  With this configuration, the parametric multi-channel decoding device of the present invention allows the all-pass filter unit constituting the reverberation signal generating means to perform filter processing by real number calculation with a low load. Can be reduced.

  Also, the parametric multi-channel decoding device of the present invention uses the following real values, that is, the filter coefficients used by the synthesis filter means:

とする構成を有している。

The configuration is as follows.

  With this configuration, the parametric multi-channel decoding device of the present invention can perform filter processing by real number calculation with a low load on the synthesis filter means, so that requirements for memory, calculation amount, and processing output can be reduced.

  In the parametric multi-channel decoding device according to the present invention, the binaural queue information includes IID (inter-channel level / intensity difference) queue information, IPD (interchannel phase / delay difference) queue information, and ICC (interchannel coherence / correlation) queue. It has a configuration that includes information.

  With this configuration, the parametric multi-channel decoding device of the present invention can restore a high-quality multi-channel signal from the downmix signal.

  In the parametric multi-channel decoding device according to the present invention, when the reverberation signal generation unit generates a reverberation signal uncorrelated with the subband signal, the ICC included in the processing coefficient generated by the interpolation unit. It has a configuration using queue information.

  With this configuration, the parametric multi-channel decoding device of the present invention can appropriately generate an uncorrelated reverberation signal.

  The parametric multi-channel decoding device of the present invention includes an analysis filter unit that converts a downmix signal into a plurality of subband signals using an analysis filter coefficient, and an interpolation that generates a processing coefficient by performing linear interpolation on binaural cue information And an all-pass filter that generates an uncorrelated reverberation signal with the sub-band signal using an all-pass filter coefficient based on the sub-band signal converted by the analysis filter means and the processing coefficient generated by the interpolation means A reverberation signal generating means, a signal distribution means for generating a multichannel signal by mixing the subband signal and the reverberation signal generated by the reverberation signal generating means based on the processing coefficient, and a synthesis filter coefficient Convert the multi-channel signal from subband signal to time signal using And the synthesis filter coefficient used by the analysis filter means, the all-pass filter coefficient used by the all-pass filter unit forming the reverberation signal generation means, and the synthesis filter coefficient used by the synthesis filter means are selectively selected. And a filter coefficient selection means for switching to a complex value or a real value.

  With this configuration, the parametric multi-channel decoding device according to the present invention enables filter coefficient selection to selectively switch the filter coefficients used by the analysis filter means, the all-pass filter section constituting the reverberation signal generation means, and the synthesis filter means to complex values and real values, respectively. Since the filter coefficient selection means is operated, the filter coefficient used by the analysis filter means, the all-pass filter part constituting the reverberation signal generation means, and the filter coefficient used by the synthesis filter part means are all complex values and complex arithmetic is executed. Low-quality mode and low-computation amount with one or more of the filter coefficients used by the analysis filter means, reverberation signal generation means, and synthesis filter means as real values by operating the filter coefficient selection means Support more than one output mode Since it is, for example, when the battery output is high if you want to save the high-quality mode, the user is output can be switched in response to the low power mode and applications.

  The parametric multi-channel decoding method of the present invention includes an analysis filter step for converting a downmix signal into a plurality of subband signals using an analysis filter coefficient, and an interpolation for generating a processing coefficient by performing linear interpolation on binaural cue information And an all-pass filter coefficient that generates a reverberation signal uncorrelated with the sub-band signal based on the sub-band signal converted by the analysis filter step and the processing coefficient generated by the interpolation step. A reverberation signal generation step including filter processing, a signal distribution step of generating a multichannel signal by mixing the reverberation signal generated by the subband signal and the reverberation signal generation step based on the processing coefficient, and a synthesis filter The multi-channel signal is supported using a coefficient. A parametric multi-channel decoding method including a synthesis filter step for converting a band signal into a time signal, wherein at least one of the analysis filter coefficient, the all-pass filter coefficient, and the synthesis filter coefficient is a real value. is doing.

  With this configuration, the parametric multi-channel decoding method of the present invention can perform filtering by real number calculation with low load by reducing complex calculation with large load when decoding a multi-channel signal. In a device with a limited battery capacity, such as broadcasting with a stereo signal downmixed to a monaural signal and transmitted / received, the requirements for memory, computation, and processing output can be reduced.

  The present invention provides a parametric multi-channel decoding apparatus and method capable of reducing memory, calculation amount, and processing output as compared with a conventional stereo decoding unit.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

(First embodiment)
A parametric multi-channel decoding device according to a first embodiment of the present invention will be described with reference to FIGS.

  As shown in FIG. 1, the parametric multi-channel decoding device according to the present embodiment includes a real number analysis filter unit 100, an interpolation unit 104, a real number reverberation signal generation unit 106, a signal distribution unit 108, and a real number synthesis filter unit. 112.

The real number analysis filter unit 100 is composed of a QMF (Quadrature Mirror Filter) filter bank, and is composed of 64 subband (frequency band) signals each having 32 samples of a monaural signal M ′ which is a restored time signal. Convert to subband signal. The real number analysis filter unit 100 constitutes analysis filter means. Here, the real analysis filter unit 100 complex valued filter coefficients Mri conventional complex analysis filter unit 800 is used (k, n) and _{G qi} ^p real-valued filter coefficients Mrr (k, n) on behalf of and _{G qr} ^p Uses to execute processing at high speed. A complex-valued filter coefficient Mri (k, n), and real-valued filter coefficients Mrr (k, n), the complex value filter coefficients _{G qi} ^p, real-valued filter coefficients _{G qr} ^p is expressed by the following equation.

  Since the high-speed processing of the real number analysis filter unit 100 described above is well known in the signal processing literature, the description thereof is omitted.

  The interpolation unit 104 generates a processing coefficient by performing linear interpolation before actually applying the binaural queue BCInfo. The interpolation unit 104 constitutes an interpolation unit. The interpolation unit 104 executes processing independently of the signal type (complex type or real number type). In the present embodiment, the binaural queue BCInfo is extracted based on the subband signals of the left channel signal L and the right channel signal R obtained by conversion using a QMF (Quadrature Mirror Filter) or the like.

The real number reverberation signal generation unit 106 is based on the monaural signal S _k (z) generated by processing by the real number analysis filter unit 100 and the ICC cue information included in the binaural cue BCInfo interpolated by the interpolation unit 104. A reverberation signal D _k (z) uncorrelated with S _k (z) is generated. The real number reverberation signal generation unit 106 constitutes a real number reverberation signal generation unit. The real reverberation signal generation unit 106 has an all-pass filter. Here, the real number reverberation signal generation unit 106 replaces the complex value all-pass filter coefficient H _ki (z) used by the conventional complex reverberation signal generation unit 806 as the filter coefficient of the all-pass filter with the real value all-pass filter coefficient H _kr (z). To use. The complex value all-pass filter coefficient H _ki (z) and the real value all-pass filter coefficient H _kr (z) are expressed by the following equations.

Based on the IID queue information and ICC queue information included in the binaural queue BCInfo interpolated by the interpolation unit 104, the signal distribution unit 108 performs real reverberation on the monaural signal S _k (z) processed by the real number analysis filter unit 100. The reverberation signal D _k (z) generated by the signal generation unit 106 is added and distributed to generate the left channel signal L _k (Z) and the right channel signal R _k (Z). This process is the most important step in the reproduction of the sound image in the present embodiment, and the signal type varies depending on the complex type or the real number type. The signal distribution unit 108 constitutes a signal distribution unit.

A function executed by the signal distribution unit 108 will be described with reference to FIG. The ratio of adding the reverberation signal D _k (z) to the monaural signal S _k (z) processed by the real number analysis filter unit 100 is determined based on the ICC queue information, and the reverberation signal D _k (z) is determined as the monaural signal S _k. The ratio distributed to the left channel signal L ′ and the right channel signal R ′ in addition to (z) is determined based on the IID queue information.

  In the parametric multichannel decoding apparatus according to the present embodiment, the additional coefficient for adding the reverberation signal is calculated by the following equation.

Here, c ₁ (b) is a gain coefficient for the right channel, and c ₂ (b) is a gain coefficient for the left channel, both of which are calculated from IID. h ₁₁ (b) is an addition coefficient when the monaural signal S _k (z) is added to the left channel signal, and h ₁₂ (b) is an addition coefficient when the monaural signal S _k (z) is added to the right channel signal. H ₂₁ (b) is an addition coefficient when the reverberation signal D _k (z) is added to the left channel signal, and h ₂₂ (b) is the addition coefficient of the reverberation signal D _k (z) to the right channel. Is an additional coefficient. As shown in FIG. 2, multiplied by the left channel signal L _{k (Z)} h ₂₁ to h ₁₁ (b) multiplied by h ₁₁ and the reverberation signal D _{k (z)} to the monaural signal S _{k (z) (b)} It is generated by adding h ₂₁ was, right channel signal R _{k (Z)} h ₂₂ to h ₁₂ (b) multiplied by h ₁₂ and the reverberation signal D _{k (z)} to the monaural signal S _{k (z) (b)} Generated by adding h ₂₂ multiplied by.
In the present embodiment, the signal distribution unit 108 calculates the coefficient for adding the reverberation signal using the above-described formula, but the present invention is not limited to this, and represents the ratio for adding the reverberation signal. Other formulas may be applied as long as the additional coefficient is calculated.

  For example, the additional coefficient can be derived by modeling the relationship between the left channel signal, the right channel signal, and the monaural signal. With reference to FIG. 3, a method for obtaining an additional coefficient by modeling will be described.

  FIG. 3 shows a vector space. When the left channel signal, the right channel signal, and the monaural signal are modeled as vectors represented by L, R, and M, the following relationship is established.

  Further, the angle θ (b) formed by L and R is expressed by the following equation.

The angle θ _L (b) formed by the left channel signal L and the monaural signal M and the angle θ _R (b) formed by the right channel signal R and the monaural signal M are expressed by the following equations.

  The real number synthesis filter unit 112 includes a filter bank, and converts the left channel signal L ′ and the right channel signal R ′ from a subband signal to a time signal. The real number synthesis filter unit 112 constitutes synthesis filter means.

  The real number synthesis filter unit 112 is composed of a synthesis QMF (Quadrature Mirror Filter) filter bank. Here, the real number synthesis filter unit 112 uses the real value filter coefficient Nrr (k, n) in place of the number complex value filter coefficient Nri (k, n) used by the conventional complex synthesis filter unit 812 for high-speed processing. Run with. The complex value filter coefficient Nri (k, n) and the real value filter coefficient Nrr (k, n) are expressed by the following equations.

  The high-speed processing of the real number synthesis filter unit 112 described above is well known in the signal processing literature, and thus the description thereof is omitted.

  The real number analysis filter unit 100, the interpolation unit 104, the real number reverberation signal generation unit 106, the signal distribution unit 108, and the real number synthesis filter unit 112 may be configured by, for example, a DSP (Digital Signal Processor).

  Hereinafter, the operation of the parametric multichannel decoding apparatus according to the present embodiment will be described.

First, the real number analysis filter unit 100 converts a monaural signal M ′, which is a time signal restored using the above-described real value filter coefficients, into sub-band (frequency band) signals each having 32 samples. The signal is converted into a band signal S _k (Z) (step S1). Next, the interpolation unit 104 performs linear interpolation on the binaural queue BCInfo to generate a processing coefficient (step S2). The real number reverberation signal generation unit 106 is based on the monaural signal S _k (z) generated by processing by the real number analysis filter unit 100 and the ICC cue information included in the binaural cue BCInfo interpolated by the interpolation unit 104. A reverberation signal D _k (z) uncorrelated with S _k (z) is generated (step S3). The real reverberation signal generation unit 106 has an all-pass filter. Here, the real number reverberation signal generation unit 106 executes processing using the above-described real value filter coefficient as the filter coefficient of the all-pass filter.

Based on the IID queue information and ICC queue information included in the binaural queue BCInfo interpolated by the interpolation unit 104, the signal distribution unit 108 performs real reverberation on the monaural signal S _k (z) processed by the real number analysis filter unit 100. The reverberation signal D _k (z) generated by the signal generation unit 106 is added to generate the left channel signal L _k (Z) and the right channel signal R _k (Z) (step S4).

The real number synthesis filter unit 112 converts the left channel signal L _k (Z) and the right channel signal R _k (Z) from the subband signal to the time signals L ′ and R ′ (step S5). The real number synthesis filter unit 112 performs processing using the above-described real value filter coefficients.

  As described above, when decoding a stereo signal, the parametric multi-channel decoding device according to the present invention can perform a filtering process based on a real number calculation with a low load by reducing a complex number calculation with a large load. It is possible to reduce the amount of calculation and processing output, and for example, it can be configured by a DSP (Digital Signal Processor) or the like, so that the chip size can be reduced.

  In the present embodiment, for ease of explanation, real value filter coefficients are used for all of the real number analysis filter unit 100, the all-pass filter unit constituting the real number reverberation signal generation unit 106, and the real number synthesis filter unit 112. However, one or two of the filter coefficient used by the real number analysis filter unit 100, the filter coefficient used by the all-pass filter unit constituting the real number reverberation signal generation unit 106, and the filter coefficient used by the real number synthesis filter unit 112 are real values. It is good.

(Second Embodiment)
FIG. 5 shows a parametric multi-channel decoding device according to the second embodiment of the present invention. In addition, in each configuration of the parametric multichannel decoding device according to the second embodiment of the present invention, the same components as those of the parametric multichannel decoding device according to the first embodiment of the present invention are the same. The description will be omitted.

  As shown in FIG. 5, the parametric multi-channel decoding apparatus according to the present embodiment includes a filter coefficient selection unit 201, a real number analysis filter unit 100, a real number reverberation signal generation unit 106, and a real number synthesis filter unit 112 (see FIG. 1). Instead, the analysis filter unit 200, the reverberation signal generation unit 206, and the synthesis filter unit 212 are the same as those provided in the parametric multichannel decoding device (see FIG. 1) according to the first embodiment.

  The filter coefficient selection unit 201 selectively outputs real-value or complex-value filter coefficients to the analysis filter unit 200, the reverberation signal generation unit 206, and the synthesis filter unit 212. The filter coefficient selection unit 201 constitutes filter coefficient selection means.

  The configuration of the analysis filter unit 200 is the same as the configuration in which the real number analysis filter unit 100 includes a function that can selectively use complex values and real values as filter coefficients. The analysis filter unit 200 constitutes an analysis filter unit.

  The configuration of the reverberation signal generation unit 206 is such that the real reverberation signal generation unit 106 has a function that allows the all-pass filter unit included in the real number reverberation signal generation unit 106 to selectively use complex values and real values as filter coefficients. The configuration is the same. The reverberation signal generation unit 206 constitutes reverberation signal generation means.

  The configuration of the synthesis filter unit 212 is the same as the configuration of the real number synthesis filter unit 112 having a function capable of selectively using complex values and real values as filter coefficients. The synthesis filter unit 212 constitutes synthesis filter means.

  Next, the operation of the parametric multichannel decoding apparatus according to this embodiment will be described.

  The filter coefficient selection unit 201 selectively outputs real-value or complex-value filter coefficients to the analysis filter unit 200, the reverberation signal generation unit 206, and the synthesis filter unit 212.

The analysis filter unit 200 converts the monaural signal M ′, which is a time signal restored using the filter coefficient selectively output by the filter coefficient selection unit 201, into 64 subbands (frequency bands) each having 32 samples. ) Is converted into a subband signal S _k (Z) consisting of signals. Next, the interpolation unit 104 performs linear interpolation on the binaural queue BCInfo to generate a processing coefficient. Reverberation signal generator 206, monaural signal _{S k} based on the ICC cue information contained in both ears queue BCInfo interpolated by monaural signal _S k (z) and an interpolation unit 104 which is generated and processed by the analysis filter unit 200 A reverberation signal D _k (z) uncorrelated with (z) is generated. The reverberation signal generation unit 206 has an all-pass filter. Here, the reverberation signal generation unit 206 performs processing using the filter coefficient selectively output by the filter coefficient selection unit 201 as the filter coefficient of the all-pass filter.

Based on the IID queue information and the ICC queue information included in the binaural queue BCInfo interpolated by the interpolation unit 104, the signal distribution unit 108 generates a reverberation signal for the monaural signal S _k (z) processed by the analysis filter unit 200. The reverberation signal D _k (z) generated by the unit 206 is added to generate the left channel signal L _k (Z) and the right channel signal R _k (Z).

The synthesis filter unit 212 converts the left channel signal L _k (Z) and the right channel signal R _k (Z) from the subband signals to time signals L ′ and R ′. The real number synthesis filter unit 212 executes processing using the filter coefficients selectively output by the filter coefficient selection unit 201.

  The parametric multi-channel decoding apparatus according to the present embodiment selectively switches the filter coefficients used by analysis filter section 200, all-pass filter section constituting reverberation signal generation section 206, and synthesis filter section 212 to complex values and real values, respectively. Since the filter coefficient selection unit 201 is provided, all the filter coefficients used by the analysis filter unit 200, the all-pass filter unit constituting the reverberation signal generation unit 206, and the synthesis filter unit 212 by operating the filter coefficient selection unit 201 are complex values. High-quality mode for performing complex number operations, and one or more filter coefficients used by the analysis filter unit 200, the all-pass filter unit constituting the reverberation signal generation unit 206, and the synthesis filter unit 212 by operating the filter coefficient selection unit 201. Low output that executes low complexity as a real value Can support two or more modes of chromatography de, for example, when the battery output is high if you want high-quality mode, the user saves the output can be switched in response to the low power mode and applications. Even if real-valued filter coefficients are used for any of the analysis filter unit 200, the all-pass filter unit constituting the reverberation signal generation unit 206, and the synthesis filter unit 212, a slight decrease in sound quality caused by this is brought about by the present invention. Considering the memory, the amount of calculation, the reduction of processing output, and the significant reduction of the chip size, it is acceptable within the low bit rate application.

  As described above, the parametric multi-channel decoding device according to the present invention can reduce complex processing without changing the syntax of the bitstream by the parametric stereo coding technique based on MPEG-4 Audio Extension 2. is there. Further, by operating the filter coefficient selection unit, for example, when the battery output is high, the high-quality mode can be switched, and when the user wants to save the output, the low-power mode can be switched according to the use. The slight degradation in sound quality that occurs when operating in the low power mode is acceptable, especially in low bit rate applications, taking into account the significant reduction in processing and chip size provided by the present invention.

  In the above-described embodiment, the binaural cue converts the left channel signal and the right channel signal into subband signals using a QMF (Quadrature Mirror Filter) or the like, and subchannels of the left channel signal L and the right channel signal R thus obtained are obtained. Although extraction is performed based on band signals, the present invention is not limited to this. Spectral conversion is performed by averaging spectral information of the left channel signal and right channel signal, and the left channel signal and right channel signal are extracted. The binaural cues may be calculated by comparing channel signal information based on spectrum bands.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention. The present invention can also be applied to a bit stream using complex values. Although the present embodiment has been described assuming a two-channel system for easy understanding, the parametric stereo coding technique of the present invention can be applied to multi-channel coding without difficulty, for example, 5 It is applicable to multi-channel systems such as 1 channel. The present invention is also applicable to home theater systems, car audio systems, electronic game systems, and the like.

  As described above, the image coding apparatus according to the present invention has an effect that it is possible to determine a compression parameter in less time than a conventional image coding apparatus, and configures a moving image. This is useful as an image encoding device for compressing and encoding an image to be compressed.

1 is a block diagram of a parametric multi-channel decoding device according to a first embodiment of the present invention. The figure explaining the function of the signal distribution part which comprises the parametric multichannel decoding apparatus shown in FIG. The figure explaining the acquisition method of the additional coefficient by the modeling applicable to the signal distribution part shown in FIG. Flowchart of a multi-channel decoding method executed by the parametric multi-channel decoding device shown in FIG. Block diagram of a parametric multi-channel decoding device according to a second embodiment of the present invention Diagram explaining the principle of conventional parametric stereo coding Block diagram of a stereo decoding unit constituting conventional parametric stereo coding

Explanation of symbols

100 Real number analysis filter section (analysis filter means)
104 Interpolation unit (interpolation means)
106 Real number reverberation signal generator (reverberation signal generator)
108 Signal distribution part (signal distribution means)
112 Real number synthesis filter unit (synthesis filter means)
200 Analysis filter section (analysis filter means)
201 Filter coefficient selection unit (filter coefficient selection means)
206 Reverberation signal generator (reverberation signal generator)
212 synthesis filter unit (synthesis filter means)
700 downmix unit 702 binaural cue extraction unit 704 audio signal encoding unit 706 multiplexing unit 708 demultiplexing unit 710 audio signal decoding unit 712 stereo decoding unit 800 complex analysis filter unit 804 interpolation unit 806 complex reverberation signal generation unit 808 signal Distribution unit 812 Complex synthesis filter unit

Claims (8)

Analysis filter means for converting the downmix signal into a plurality of subband signals using the analysis filter coefficient;
Interpolation means for performing linear interpolation on binaural queue information to generate processing coefficients;
An all-pass filter unit that generates a reverberation signal uncorrelated with the sub-band signal using an all-pass filter coefficient based on the sub-band signal converted by the analysis filter unit and the processing coefficient generated by the interpolation unit; Reverberation signal generating means;
Signal distribution means for generating a multichannel signal by mixing the subband signal and the reverberation signal generated by the reverberation signal generation means based on the processing coefficient;
Synthesis filter means for converting the multichannel signal from a subband signal to a time signal using a synthesis filter coefficient;
Any one or more of the analysis filter coefficient used by the analysis filter unit, the all-pass filter coefficient used by the all-pass filter unit forming the reverberation signal generation unit, and the synthesis filter coefficient used by the synthesis filter unit are real values. A parametric multi-channel decoding device. The filter coefficients used by the analysis filter means are the following real values:

and

The parametric multi-channel decoding device according to claim 1, wherein The all-pass filter coefficient used by the all-pass filter unit forming the reverberation signal generating unit is expressed by the following real value, that is,

The parametric multi-channel decoding device according to claim 1, wherein The complex value filter coefficients used by the synthesis filter means are the following real values:

The parametric multi-channel decoding device according to claim 1, wherein The binaural queue information includes IID (inter-channel level / intensity difference) queue information, IPD (interchannel phase / delay difference) queue information, and ICC (interchannel coherence / correlation) queue information. The parametric multi-channel decoding device described. 6. The reverberation signal generation unit uses the ICC queue information included in the processing coefficient generated by the interpolation unit when generating a reverberation signal uncorrelated with the subband signal. Parametric multi-channel decoding device according to 1. Analysis filter means for converting the downmix signal into a plurality of subband signals using the analysis filter coefficient;
Interpolation means for performing linear interpolation on binaural queue information to generate processing coefficients;
An all-pass filter unit that generates a reverberation signal uncorrelated with the sub-band signal using an all-pass filter coefficient based on the sub-band signal converted by the analysis filter unit and the processing coefficient generated by the interpolation unit; Reverberation signal generating means;
Signal distribution means for generating a multichannel signal by mixing the subband signal and the reverberation signal generated by the reverberation signal generation means based on the processing coefficient;
Synthesis filter means for converting the multi-channel signal from a subband signal to a time signal using a synthesis filter coefficient;
The analysis filter coefficient used by the analysis filter means, the all-pass filter coefficient used by the all-pass filter section forming the reverberation signal generation means, and the synthesis filter coefficient used by the synthesis filter means are selectively converted into complex values or real values. A parametric multi-channel decoding device comprising: a filter coefficient selection unit for switching. An analysis filter step for converting the downmix signal into a plurality of subband signals using the analysis filter coefficients;
An interpolation step for performing linear interpolation on binaural cue information to generate processing coefficients;
An all-pass filter process using an all-pass filter coefficient that generates a reverberation signal uncorrelated with the sub-band signal based on the processing coefficient generated by the sub-band signal converted by the analysis filter step and the interpolation step. A reverberation signal generation step;
A signal distribution step of generating a multichannel signal by mixing the subband signal and the reverberation signal generated by the reverberation signal generation step based on the processing coefficient;
A parametric multi-channel decoding method including a synthesis filter step of converting the multi-channel signal from a subband signal to a time signal using a synthesis filter coefficient;
One or more of the analysis filter coefficient, the all-pass filter coefficient, and the synthesis filter coefficient are real values, and the parametric multichannel decoding method.

Images