KR101176703B1 - Decoder and decoding method for multichannel audio coder using sound source location cue - Google Patents

Abstract

Disclosed are a multi-channel audio decoding apparatus and method based on sound source location clues. A multi-channel audio decoding apparatus based on a sound source position clue includes: a demultiplexer which receives a signal and parses the received signal into an audio bitstream and an additional information bitstream; An audio decoder to restore a downmix signal based on the audio bitstream; An integrated upmixing unit for predicting a multichannel signal using the downmix signal and the side information bitstream, and generating an upmixing signal by upmixing the downmix signal based on the multichannel signal; And an integrated filter bank windowing unit configured to extract a time domain signal by performing an integrated filter bank on the upmix signal, and extract an output signal by windowing the upmix signal.

SSLCC, multichannel, downmix, Huffman coding, inverse quantization.

Description

Multi-channel audio decoding apparatus and method based on sound source location clues {DECODER AND DECODING METHOD FOR MULTICHANNEL AUDIO CODER USING SOUND SOURCE LOCATION CUE}

The present invention relates to an apparatus and method for multi-channel audio decoding based on sound source location clues.

The basic coding concept of Sound Source Location Cue Coding (SSLCC) starts from Spatial Audio Coding (SAC).

The SAC is a technique for compressing a multi-channel audio signal based on sound source location clues, and maximizes the compression efficiency by removing redundancy of each channel signal based on spatial cues perceived by a person in space. can do.

In addition, the multichannel signal is basically downmixed, so that the downmixed signal is a core signal of the transmitted audio signal. That is, the basic principle of the SAC technique is that it can be reproduced through existing stereo audio.

The SSLCC is one of such SAC schemes. It is a spatial cue recognized by a person in space, and considers the position of a sound source, extracts the position information of the sound source from a multi-channel signal, and transmits it.

In this case, since the information extracted through the SAC coding strategy has a small amount of information, the information can be transmitted to the redundant data area. When the destination receiving the information supports the SAC coding scheme, the transmitted redundancy information may be used to restore the sound source position-based multi-channel audio signal from the downmix stereo signal.

However, in order to recover the multi-channel audio signal based on sound source position cues from the downmix stereo signal using the surplus information, the same T / F conversion method is used as the T / F conversion method used in the process of extracting information through the SAC coding strategy. Since the T / F conversion method used in the information extraction through the SAC coding strategy is not the T / F conversion method optimized for the destination, it may affect the conversion process.

Therefore, there is a need for an apparatus or method capable of reconstructing a multi-channel audio signal based on a sound source position clue with a T / F conversion method optimized at a destination.

The present invention receives a multi-channel audio signal, compresses it, and compresses and transmits a stereo signal through a core stereo codec, thereby providing backward compatibility with the existing stereo audio codec and simultaneously transmitting multi-channel audio. It provides a possible decoding apparatus and method.

The present invention provides a decoding apparatus and method capable of varying T / F conversion for a stereo downmix audio signal according to an option by using a time domain alias cancellation cancellation (TDAC) file bank.

A multi-channel audio decoding apparatus based on a sound source position clue according to an embodiment of the present invention comprises: a demultiplexer for receiving a signal and parsing the received signal into an audio bitstream and an additional information bitstream; An audio decoder to restore a downmix signal based on the audio bitstream; An integrated upmixing unit for predicting a multichannel signal using the downmix signal and the side information bitstream, and generating an upmixing signal by upmixing the downmix signal based on the multichannel signal; And an integrated filter bank windowing unit configured to extract a time domain signal by performing an integrated filter bank on the upmix signal, and extract an output signal by windowing the upmix signal.

In addition, a multi-channel audio decoding method based on a sound source position clue according to an embodiment of the present invention includes parsing a received signal into an audio bitstream and an additional information bitstream; Restoring a downmix signal based on the audio bitstream; Restoring side information by decoding the side information bitstream; Predicting a multi-channel signal using the downmix signal and the side information; Generating an upmix signal by upmixing the downmix signal based on the multi-channel signal; Extracting a time domain signal by performing an integrated filter bank on the upmix signal; And windowing the upmix signal to extract an output signal.

According to the present invention, by receiving a multi-channel audio signal and compressing it and compressing and transmitting a stereo signal through a core stereo codec, a backward compatible with existing stereo audio codec and at the same time, multi-channel audio Transmission is possible.

According to the present invention, a T / F conversion for a stereo downmix audio signal may be varied according to an option by using a time domain alias cancellation cancellation (TDAC) file bank.

Hereinafter, various embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

1 is an example of an audio source location clue based multi-channel audio encoding apparatus according to an embodiment of the present invention.

The multi-channel audio encoding apparatus 100 based on sound source location cues according to an embodiment of the present invention is a 5-channel multi-channel audio encoding apparatus based on sound source location cue coding (SSLCC), as shown in FIG. 1. The unit 110, the analyzer 120, the downmix processor 130, the audio encoder 140, and the multiplexer 150 may be configured.

In this case, the multi-channel audio encoding apparatus 100 based on the sound source position clue may be extended to the multi-channel audio content based on the sound source position clue of 5 channels or more.

The preprocessing filter bank unit 110 preprocesses the multi-channel input audio signal input to the multi-channel audio encoding apparatus 100 based on the sound source position clue, and passes the pre-processed input audio signal through a filterbank. Can be changed to a frequency domain signal. In this case, the filter bank may be configured to perform time to frequency (T / F) conversion based on sub-band analysis (MDCT, MDST, DFT, etc.).

In this case, the input audio signal is an input signal LF (Left Front), an input signal RF (Right Front). The input signal C (Center Front), the input signal Ls (Left Surround), and the input signal Rs (Right Surround) may be included.

The analyzer 120 extracts a spatial cue from an input audio signal converted from the preprocessing filter bank unit 110 into a frequency domain signal, and expresses the spatial cue as an additional information bitstream. In this case, the analyzer 120 may compress the input audio signal and transmit the compressed audio signal to the downmix processor 130.

The downmix processor 130 may downmix the input audio signal compressed by the analyzer 120 in a downmixing manner of the audio signal in the frequency domain. In addition, the downmix processor 130 may downmix according to the recommendations of the ITU-T.

The downmixed audio signal in the downmix processor 130 may be represented as a bitstream through general stereo audio. In this case, as the general stereo audio, MP3 (MPEG Layer III) or AAC (Advanced Audio Coding) may be utilized.

The audio encoder 140 may encode the downmixed audio signal by the downmix processor 130.

The multiplexer 150 may combine and transmit the signal encoded by the audio encoder 140 and the side information bitstream transmitted by the analyzer 120.

2 is a diagram illustrating an audio source position clue based multi-channel audio decoding apparatus according to an embodiment of the present invention.

The multi-channel audio decoding apparatus 200 based on the sound source position clue according to the embodiment of the present invention is a SSLCC-based 5-channel multi-channel audio decoding apparatus as shown in FIG. 2, the demultiplexer 210 and the audio decoder 220. ), The windowing filter bank unit 230, the integrated upmixer 240, and the integrated filter bank windowing unit 250 may be configured.

The demultiplexer 210 may receive a signal transmitted by the multiplexer 150 and parse the received signal into an audio bitstream and an additional information bitstream.

The audio decoder 220 may restore the downmix signal based on the audio bitstream.

The windowing filter bank unit 230 may apply an analytical filter bank to the downmix signal to perform T / F conversion, and may window and transmit the T / F converted downmix signal.

The integrated upmixing unit 240 predicts a multichannel signal using the downmix signal and the side information bitstream, and upmixes the downmix signal based on the multichannel signal to generate an upmix signal. Can be generated.

In detail, the integrated upmixing unit 240 separates amplitude information and phase information from the downmix signal, weights the phase information by windowing a random sequence previously stored based on the amplitude information, and applies the weight information to the amplitude information. The multi-channel signal may be predicted based on the weighted phase information.

In this case, the integrated upmixing unit 240 may complex-convert the downmix signal and separate amplitude information and phase information from the down-converted downmix signal.

In addition, the integrated upmixing unit 240 models the spectral window for correcting the phase information based on the envelope of the amplitude information, uses the spectral window for the previously stored random sequence, and windowes the windowed window. A random sequence may be used to weight the phase information.

In addition, the integrated upmixer 240 may combine the amplitude information and the weighted phase information, and inversely complex convert the combined amplitude information and the weighted phase information to predict a multi-channel signal.

The integrated filter bank windowing unit 250 may extract a time domain signal by performing a synthesis filter bank on the upmix signal, and extract an output signal by windowing the upmix signal. .

3 illustrates an example of a multi-channel audio decoding apparatus based on a sound source location clue according to another embodiment of the present invention.

The multi-channel audio decoding apparatus 300 based on sound source position clue according to another embodiment of the present invention is a multi-channel audio decoding apparatus based on sound source position clue to which Real-Transform is applied, as shown in FIG. 2. The demultiplexer 310, the TDAC filter bank 320, the integrated upmixer 330, and the integrated filterbank windowing unit 340 may be configured.

SSLCC basically follows the DFT filterbank (transform). However, to work with core stereo audio, various filterbank types can be applied.

In this case, even if the type of the filter bank is changed, the operating principle of the synthesis of the SSLCC analyzer 120, the integrated upmixer 240, and the integrated filter bank windowing unit 250 may be the same.

That is, since a decorelator is not applied during stereo transmission, a real filterbank (transform) may be possible. The multi-channel audio decoding apparatus 300 based on the sound source location clue according to another embodiment of the present invention is an embodiment of a configuration capable of interworking with a core codec using the MDCT capable of real conversion.

The TDAC filter bank 320 restores the downmix signal based on the audio bitstream and then omits the process of applying an analyse filter bank to the downmix signal and the windowing process, and skips the integrated upmixer 330. In this case, the signal transmitted by the TDAC filter bank 320 may be a signal L whose frequency is downmixed and a signal R whose frequency is downmixed.

The integrated filterbank windowing unit 340 extracts a time domain signal by performing a synthesis filterbank on the upmixing signal generated by the integrated upmixing unit 330, and extracts the upmixing signal from the core of the stereo audio. The output signal can be extracted by windowing to map with the rushing windowing.

In this case, the demultiplexer 310 and the integrated upmixer 330 may include the demultiplexer 210 and the integrated upmixer 240 of the multi-channel audio decoding apparatus 200 based on the sound source position clue according to an embodiment of the present invention. Since the same configuration, detailed description thereof will be omitted.

The multi-channel audio decoding apparatus 300 based on the sound source position clue according to another embodiment of the present invention may vary the T / F conversion for the stereo downmix audio signal according to an option. For example, when the operation of the inverse correlator is turned off in the decoding apparatus, real T / F conversion may be applied. In this case, complex T / F conversion is possible, but phase information is not used even when performing complex T / F conversion.

However, when the operation requiring the phase information in the decoding apparatus is turned on, that is, when the inverse correlator is required, the complex T / F conversion must be used. DFT is the basis for complex T / F conversion, but MDCT / MDST can be utilized as a complex transform pair.

4 is a diagram illustrating an apparatus for decoding an additional information bitstream according to an embodiment of the present invention.

An apparatus for decoding an additional information bitstream according to an embodiment of the present invention is an apparatus for decoding a virtual sound location angle (VSLA), which is additional information, from the additional information bitstream parsed by the demultiplexer 210. As shown in FIG. Likewise, the Huffman decoder 410 and the inverse quantizer 420 may be configured.

In addition, the apparatus for decoding the additional information bitstream according to an embodiment of the present invention may be included in the windowing filter bank unit 230 or the integrated filter bank windowing unit 250.

The Huffman decoder 410 may generate a differential index by Huffman decoding the additional information bitstream using the Huffman codebook.

In this case, the Huffman decoder 410 may generate the Huffman codebook including an inverse difference coding unit 411, a difference coding unit 412, a mapping unit 413, and a Huffman coding unit 414.

In this case, the inverse difference coding unit 411 may decode the original index by performing inverse difference coding on the basis of the information on the previously processed previous frame and the Huffman coding type.

In addition, the differential coding unit 412 may generate index information by performing differential coding after removing negative information from the original index in response to the information of the sign bit.

Next, the mapping unit 413 removes the offset information for removing the negative information from the index, and then maps the index according to a frequency solution and classifies the index into other bands except for the first subband and the first subband. can do.

Finally, the Huffman coding unit 414 may generate the Huffman codebook by applying the Huffman coding method to each of the first subband and the other bands except for the first subband.

The Huffman decoder 410 may decode the Huffman by referring to the Huffman codebook of Table 1 below.

Index Num of bits Code word Index Num. of bits Codeword 0 5 0x17 16 5 0x1d One 8 0x64 17 5 0x19 2 8 0x65 18 5 0x1c 3 8 0xf0 19 5 0x16 4 8 0xf1 20 5 0x18 5 7 0x33 21 5 0x14 6 7 0x79 22 5 0x13 7 6 0x18 23 5 0x15 8 6 0x22 24 5 0x1b 9 6 0x23 25 5 0x10 10 6 0x3d 26 5 0x0e 11 5 0x0b 27 5 0x0f 12 5 0x12 28 5 0x0d 13 5 0x1a 29 5 0x0a 14 4 0x04 30 2 0x00 15 5 0x1f

In addition, the Huffman decoder 410 may decode Huffman with reference to the Huffman codebook of Table 2 when the signal received by the demultiplexer 210 is a signal quantized to 5 bits.

Index Differential Frequency Differential Frequency Index Differential
Time Differential
Time Number of bits Codeword Number of bits Codeword Number of bits Codeword Number of bits Codeword -30 34 0x7907FFFF 27 0x5D0FFF One 3 0x00000003 3 0x000003 -29 34 0x7907FFFE 27 0x5D0FFE 2 4 0x00000003 5 0x000000 -28 33 0x3C83FFFE 26 0x2E87FE 3 5 0x00000004 5 0x000004 -27 32 0x1E41FFFE 25 0x1743FE 4 6 0x00000004 6 0x000004 -26 31 0x0F20FFFE 24 0x0BA1FE 5 6 0x00000010 7 0x000004 -25 30 0x07907FFE 19 0x005D0E 6 6 0x00000013 7 0x000014 -24 29 0x03C83FFE 17 0x001742 7 7 0x00000025 8 0x000014 -23 20 0x0001E41E 15 0x0015E6 8 8 0x0000001D 9 0x000014 -22 28 0x01E41FFE 17 0x00174D 9 9 0x00000031 9 0x000054 -21 18 0x00007906 16 0x000BA0 10 9 0x0000003D 10 0x00002C -20 17 0x00003C82 15 0x0005D2 11 10 0x00000070 10 0x0000AD -19 15 0x00000F22 15 0x0005D1 12 11 0x000000C0 11 0x00005E -18 15 0x00000F21 14 0x0002EA 13 11 0x000000E3 11 0x000159 -17 13 0x000003CE 14 0x000AC4 14 12 0x000001E5 12 0x0000B8 -16 13 0x00000307 13 0x000563 15 13 0x00000306 12 0x0002B0 -15 12 0x000001E6 12 0x0000BB 16 13 0x000003C9 12 0x0002BD -14 12 0x00000182 12 0x0000B9 17 14 0x0000079E 13 0x000578 -13 11 0x000000E2 11 0x00015F 18 14 0x0000079F 14 0x0002EB -12 10 0x00000078 11 0x00005F 19 16 0x00001E47 14 0x000AC5 -11 10 0x00000061 10 0x0000AE 20 16 0x00001E40 15 0x0015E4 -10 9 0x00000039 10 0x00002D 21 17 0x00003C8C 15 0x0015E5 -9 8 0x0000001F 9 0x000055 22 17 0x00003C8D 16 0x002BCE -8 8 0x00000019 9 0x000015 23 19 0x0000F20E 17 0x00174C -7 7 0x00000024 8 0x000015 24 27 0x000F20FFE 16 0x000BA7 -6 7 0x0000000D 7 0x00000B 25 26 0x0007907FE 16 0x002BCF -5 6 0x00000011 6 0x00000B 26 25 0x0003C83FE 18 0x002E86 -4 6 0x00000005 6 0x000003 27 24 0x0001E41FE 23 0x05D0FE -3 5 0x00000005 5 0x000003 28 23 0x0000F20FE 22 0x02E87E -2 4 0x00000005 4 0x000003 29 22 0x00007907E 21 0x01743E -One 4 0x00000000 3 0x000002 30 21 0x00003C83E 20 0x00BA1E 0 One 0x00000001 One 0x000001

In addition, the Huffman decoder 410 may decode Huffman with reference to the Huffman codebook of Table 3 when the signal received by the demultiplexer 210 is a signal quantized to 4 bits.

Index Differential Frequency Differential Frequency Index Differential
Time Differential
Time Number of bits Codeword Number of bits Number of bits Number of bits Codeword Number of bits Codeword -14 21 0xFCA7F 17 0x1F9EF One 3 0x00000 3 0x00005 -13 21 0xFCA7E 17 0x1F9EE 2 4 0x00004 4 0x0000D -12 17 0x0FCA6 13 0x01FFE 3 5 0x0000C 5 0x0001D -11 16 0x07E52 13 0x01F9F 4 6 0x0001D 6 0x0003C -10 14 0x01F95 12 0x00FFE 5 7 0x0003D 7 0x0007C -9 11 0x003F3 11 0x007FE 6 8 0x0007D 8 0x000FD -8 11 0x003F0 10 0x003FE 7 9 0x000FD 9 0x001F8 -7 8 0x0007F 9 0x001FE 8 11 0x003F1 10 0x003F2 -6 8 0x0007C 8 0x000FE 9 12 0x007E4 11 0x007E6 -5 7 0x0003C 7 0x0007D 10 13 0x00FCB 12 0x00FCE -4 6 0x0001C 6 0x0003D 11 15 0x03F28 13 0x01FFF -3 5 0x0000D 5 0x0001C 12 20 0x7E53E 14 0x03F3C -2 4 0x00005 4 0x0000C 13 19 0x3F29E 15 0x07E7A -One 3 0x00001 3 0x00004 14 18 0x1F94E 16 0x0FCF6 0 One 0x00001 One 0x00000

The inverse quantization unit 420 may restore additional information by inversely quantizing the differential index into the inverse quantization table. In more detail, the inverse quantization unit 420 may inverse quantize by mapping quantization tables corresponding to each VSLA to virtual sound location angle (VSLA) information per frame. At this time, the multi-channel audio based on the sound source position clue according to the present invention is basically carried out by the frame-based DFT or MDCT, so that the smoothing between frames is mainly overlapped through windowing. Can be satisfied by the method.

In this case, the inverse quantization unit 420 may dequantize by mapping the quantization table shown in Table 4 below when the VSLA information is left half-plane vector angle (LHA).

Idx -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 LHA [idx] -55 -51.3 -47.6 -44.0 -40.3 -36.6 -33.0 -29.3 -25.6 -22.0 -18.3 Idx -4 -3 -2 -One 0 One 2 3 4 5 6 LHA [idx] -14.6 -11 -7.3 -3.6 -3.6 7.3 11 14.6 18.3 22.0 Idx 7 8 9 10 11 12 13 14 15 LHA [idx] 25.6 29.3 33.0 36.6 40.3 44.0 47.6 51.3 55

In this case, the dequantization unit 420 restores the additional information to dequantize by mapping the quantization table shown in Table 5 below when the VSLA information is a right half-plane vector angle (RHA).

Idx -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 RHA [idx] -55 -51.3 -47.6 -44.0 -40.3 -36.6 -33.0 -29.3 -25.6 -22.0 -18.3 Idx -4 -3 -2 -One 0 One 2 3 4 5 6 RHA [idx] -14.6 -11 -7.3 -3.6 -3.6 7.3 11 14.6 18.3 22.0 Idx 7 8 9 10 11 12 13 14 15 RHA [idx] 25.6 29.3 33.0 36.6 40.3 44.0 47.6 51.3 55

In this case, the dequantization unit 420 restores the additional information to dequantize by mapping the quantization table shown in Table 6 below when the VSLA information is Left Subsequent Vector Angle (LSA).

Idx -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 LSA [idx] -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 Idx -4 -3 -2 -One 0 One 2 3 4 5 6 LSA [idx] -4 -3 -2 -One 0 One 2 3 4 5 6 Idx 7 8 9 10 11 12 13 14 15 LSA [idx] 7 8 9 10 11 12 13 14 15

In this case, the inverse quantization unit 420 restores the additional information may be dequantized by mapping the quantization table shown in Table 7 below when the VSLA information is a right subsequent vector angle (RSA).

Idx -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 RSA [idx] -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 Idx -4 -3 -2 -One 0 One 2 3 4 5 6 RSA [idx] -4 -3 -2 -One 0 One 2 3 4 5 6 Idx 7 8 9 10 11 12 13 14 15 RSA [idx] 7 8 9 10 11 12 13 14 15

In addition, the inverse quantization unit 420 may extract parameter values satisfying Equation 1 from VSLA information obtained from each index information.

In this case, the number of subbands may be obtained from bsFreqRes, which is a subband number defining variable included in the additional information, and the dequantization unit 420 may also map the number of VSLA information transmitted according to the number of subbands. In addition, the maximum number of bands may be changed by varying the resolution according to the frequency characteristics of the bit rate or frame based on 28 bands (M _par = 28).

Inverse quantization unit 420 may be mapped as shown in Table 8 through the mapsubbands (bsFreqRes, M _{par )} .

m Mapfunc (m, M _{par )} M _par = 28 M _par = 20 M _par = 14 M _par = 10 M _par = 7 M _par = 5 M _par = 4 0 0 0 0 0 0 0 0 One One One 0 0 0 0 0 2 2 2 One One 0 0 0 3 3 3 One One 0 0 0 4 4 4 2 2 One One 0 5 5 5 3 2 One One 0 6 6 6 4 3 2 One One 7 7 7 4 3 2 One One 8 8 8 5 4 2 2 One 9 9 9 6 4 3 2 One 10 10 10 6 5 3 2 One 11 11 11 7 5 3 2 2 12 12 12 7 6 3 2 2 13 13 13 8 6 4 2 2 14 14 14 8 7 4 3 2 15 15 14 8 7 4 3 2 16 16 15 9 7 4 3 2 17 17 15 9 7 4 3 2 18 18 16 10 8 5 3 2 19 19 16 10 8 5 3 2 20 20 17 11 8 5 3 2 21 21 17 11 8 5 3 2 22 22 18 12 9 6 4 3 23 23 18 12 9 6 4 3 24 24 18 12 9 6 4 3 25 25 19 13 9 6 4 3 26 26 19 13 9 6 4 3 27 27 19 13 9 6 4 3

The inverse quantization unit 420 may process Equation 1 using the resolution of each band designed based on the ERB band, and when M _par = 28 in the case of mapping as shown in Table 8, the resolution of each band is It may be as shown in Table 9 below.

m M _par = 28 kHz 0 0 0.0702 One One 0.1639 2 2 0.2576 3 3 0.3512 4 4 0.4449 5 5 0.5385 6 6 0.6322 7 7 0.7259 8 8 0.9132 9 9 1.1005 10 10 1.2878 11 11 1.4751 12 12 1.8498 13 13 2.2244 14 14 2.599 15 15 2.9737 16 16 3.7229 17 17 4.4722 18 18 5.2215 19 19 5.9707 20 20 6.72 21 21 7.4693 22 22 8.5932 23 23 9.7171 24 24 11.2156 25 25 13.0888 26 26 15.3366 27 27 24

In this case, the cyclic prefix remover 210, the inverse discrete Fourier transform 220, and the guard band remover 230 may be configured to have the same number as the number of receive antennas used in the decoding apparatus and correspond to each receive antenna. have.

After the windowing filter bank unit 230 basically performs T / F conversion, the windowing filter bank unit 230 may define a single processing band by combining bands in a frequency domain.

For example, when performing a 2048-point DFT transformation, bins in a frequency domain may be defined as a single processing band by centering the positions of start and stop bins as shown in Table 10 below.

m 111 = 28 Frequncy bin Start bin's
position End bin's position 0 0 One 3 One One 4 7 2 2 8 11 3 3 12 15 4 4 16 19 5 5 20 23 6 6 24 27 7 7 28 31 8 8 32 39 9 9 40 47 10 10 48 55 11 11 56 63 12 12 64 79 13 13 80 95 14 14 96 111 15 15 112 127 16 16 128 159 17 17 160 191 18 18 192 223 19 19 224 255 20 20 256 287 21 21 288 319 22 22 320 367 23 23 368 415 24 24 416 479 25 25 480 559 26 26 560 655 27 27 656 1025

In this case, the integrated upmixing unit 240 may restore the multi-channel signal through sound phase in the subband of each channel of the downmix signal based on the VSLA information recovered from the additional information bitstream. In detail, as shown in FIG. 5 described below, the subband signal of each channel can be predicted by predicting and applying power information in each subband from a panning angle.

5 is an example of a process of estimating the gain of each channel by the integrated upmixing unit according to an embodiment of the present invention.

As illustrated in FIG. 5, the integrated upmixing unit 240 may reconstruct the sound information of each channel hierarchically and predict the gain of each channel.

First, the integrated upmixer 240 may restore the LHA [idx] 510 and the RHA [idx] 520.

Next, the integrated upmixer 240 predicts the gLs [idx] 530 from the LHA [idx] 510, restores the LSA [idx] 511, and gRs [from the RHA [idx] 520. idx] 540 may be predicted and the RSA [idx] 521 may be restored.

The integrated upmixer 240 then predicts gL [idx] 550 and gCL [idx] 512 from the LSA [idx] 511 and gRs [idx] from the RSA [idx] 521. 560 and gCR [idx] 522 may be predicted.

Finally, the integrated upmixer 240 may predict gCL [idx] / sqrt (2) 570 from gCL [idx] 512 and gCR [idx] 522. In this case, gCL [idx] / sqrt (2) may be a value obtained by scaling the gain of the center channel as gCL [idx] 512 * 0.7071.

The integrated upmixing unit 240 may generate an upmixing signal by upmixing the downmix signal based on the multi-channel signal predicted through the above process.

In this case, when X _dmxL (m, k) is the k-th frequency bin of the m-th subband of the transmitted Left downmix signal, 'Left upmixing matrix' may satisfy Equation 2 described below.

In addition, 'Rightupmixing Matrix' for the Right downmix signal may satisfy Equation 3 below.

In addition, the integrated upmixing unit 240 may include D _L and D _R , which are decorrelators based on DFT.

The D _L and D _R may operate in a high-complexity mode and a low-complexity mode in a normal mode. In this case, the D _L and the D _R may be generated only in the decoder, and may operate in a high complexity mode when generating a high sound feeling, and may operate in a low complexity mode when generating general sound quality.

The D _L and D _R generate an inverse correlation signal by performing a matrixing operation of Equation 4 below for L (m, k) and R (m, k) in the high complexity mode. can do.

In addition, D _L and D _R satisfy the following Equation 5 in the normal mode and may not generate an inverse correlation signal.

In addition, the integrated upmixing unit 240 may perform upmixing by using the values calculated in Equation 2 and Equation 3 below.

In this case, α (m) may be a factor representing a correlation between L and R signals for each band. In addition, δ is a fixed coefficient and may be a fixed coefficient which is an inverse of a mixing coefficient of a surround signal when the mixer is downmixed.

In this case, α (m) may be obtained by using the values calculated in Equations 4 and 5 in Equation 7 below.

이며,

는 0≤

≤1의 범위 내에서 정의될 수 있다.In this case, λ is a weighting coefficient and may be a value for adjusting the degree of mixing of the decorrelation signal. Therefore, α (m) is 0 ≦ α (m) ≦

Is,

Is 0≤

It can be defined within the range of ≤ 1.

Also, the _wet L (m, k) and the _wet R (m, k) may be generated through decorrelation processing performed in the decorrelator as a decorrelative signal.

6 illustrates an example of a decorrelator according to an embodiment of the present invention.

The decorrelator 600 according to an embodiment of the present invention is included in the integrated upmixing unit 240 to generate a decorrelative signal, as shown in FIG. 6, the complex converter 610 and the amplitude information extractor. 620, the phase information extractor 630, the random sequence storage unit 640, the windowing unit 650, the phase shifter 660, the integrator 670, and the inverse complex transform unit 680. Can be.

The complex converter 610 may complex convert the downmix signal.

The amplitude information extractor 620 and the phase information extractor 630 may extract amplitude information and phase information from the downmix signal converted by the complex converter 610 to separate the downmix signal.

The windowing unit 650 models a spectral window for correcting phase information based on the envelope of the amplitude information extracted by the amplitude information extracting unit 620, and stores the spectral window in the random sequence stored in the random sequence storage unit 640. Windowing can be done using spectral windows.

In this case, the number of random sequences previously stored in the random sequence storage unit 640 is determined according to the number of the downmix signals. That is, to generate the _wet L (m, k) and _wet R (m, k). Different random sequences are used, and the correlation between the two random sequences used may be close to zero.

The phase converter 660 may weight the phase information extracted by the phase information extractor 630 using the random sequence windowed by the windower 650.

The integrator 670 may combine the amplitude information extracted by the amplitude information extractor 620 and the phase information weighted by the phase converter 660.

The inverse complex transform unit 680 may calculate an inverse correlation signal by inverse complex transforming the information combined in the integrator 670.

7 is a flowchart illustrating a multi-channel audio decoding method based on a sound source location clue according to an embodiment of the present invention.

In operation S710, the demultiplexer 210 may receive a signal transmitted by the multiplexer 150 and parse the received signal into an audio bitstream and an additional information bitstream.

In operation S720, the audio decoder 220 may restore a downmix signal based on the audio bitstream parsed in operation S710.

In operation S730, the Huffman decoder 410 may generate a differential index by Huffman decoding the additional information bitstream parsed in operation S710 by using the Huffman codebook.

In operation S740, the inverse quantization unit 420 may restore additional information by inversely quantizing the differential index generated in operation S710 with an inverse quantization table. In more detail, the inverse quantization unit 420 may inverse quantize by mapping quantization tables corresponding to each VSLA to virtual sound location angle (VSLA) information per frame.

In operation S750, the integrated upmixer 240 predicts a multi-channel signal by using the downmix signal reconstructed in operation S720 and the additional information reconstructed in operation S740. Upmixing the downmix signal may be performed to generate an upmix signal.

In operation S760, the integrated filter bank windowing unit 250 performs an integrated filter bank on the upmixing signal generated in operation S750 to extract a time domain signal, and windows the upmixing signal to output an output signal. Can be extracted.

As described above, the multi-channel audio decoding apparatus and method based on the sound source position clue according to the present invention receives a multi-channel audio signal, compresses it, and compresses and transmits the stereo signal through a core stereo codec. It provides backward compatibility with existing stereo audio codecs and enables multichannel audio transmission.

In the present invention as described above has been described by the specific embodiments, such as specific components and limited embodiments and drawings, but this is provided to help a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations are possible from these descriptions.

Therefore, the spirit of the present invention should not be construed as being limited to the described embodiments, and all of the equivalents or equivalents of the claims, as well as the following claims, are included in the scope of the present invention.

1 is an example of an audio source location clue based multi-channel audio encoding apparatus according to an embodiment of the present invention.

2 is a diagram illustrating an audio source position clue based multi-channel audio decoding apparatus according to an embodiment of the present invention.

3 illustrates an example of a multi-channel audio decoding apparatus based on a sound source location clue according to another embodiment of the present invention.

4 is a diagram illustrating an apparatus for decoding an additional information bitstream according to an embodiment of the present invention.

5 is an example of a process of estimating the gain of each channel by the integrated upmixing unit according to an embodiment of the present invention.

6 illustrates an example of a decorrelator according to an embodiment of the present invention.

7 is a flowchart illustrating a multi-channel audio decoding method based on a sound source location clue according to an embodiment of the present invention.

Claims (12)

A demultiplexer for receiving a signal and parsing the received signal into an audio bitstream and an additional information bitstream; An audio decoder for restoring a downmix signal based on the audio bitstream and transmitting the downmix signal to an integrated upmixing unit without applying an analytical filter bank to the restored downmix signal; An integrated upmixing unit for predicting a multichannel signal using the downmix signal and the side information bitstream, and generating an upmixing signal by upmixing the downmix signal based on the multichannel signal; An integrated filter bank windowing unit extracts a time domain signal by performing an integrated filter bank on the upmix signal, and extracts an output signal by windowing the upmix signal. Apparatus for decoding multi-channel audio based on the sound source position clues, comprising: a. The method of claim 1, And the audio decoder comprises a stereo core codec using a TDAC (bank of time domain alias cancellation) files. The method of claim 1, The integrated filter bank windowing unit windowing the upmixed signal to be mapped with the analytical windowing of the core stereo audio, multi-channel audio decoding apparatus based on the sound source position. The method of claim 1, The integrated upmixing unit separates amplitude information and phase information from the downmix signal, weights the phase information by windowing a pre-stored random sequence based on the amplitude information, and adds the amplitude information and weighted values. A multi-channel audio decoding apparatus based on sound source position clues, comprising: predicting a multi-channel signal based on phase information. 5. The method of claim 4, And the integrated upmixing unit complex-converts the downmix signal and separates amplitude information and phase information from the complex-converted downmix signal. 5. The method of claim 4, The integrated upmixer models a spectral window for correcting phase information based on the envelope of the amplitude information, windowes the spectral window using a pre-stored random sequence, and uses a windowed random sequence. And weighting the phase information by using the multi-channel audio decoding apparatus based on the sound source position. 5. The method of claim 4, The integrated upmixing unit combines the amplitude information and the weighted phase information, and inversely complex-converts the combined amplitude information and the weighted phase information to predict a multi-channel signal. Multi-channel audio decoding device. Parsing the received signal into an audio bitstream and an additional information bitstream; Restoring a downmix signal that does not apply an analytical filter bank based on the audio bitstream; Generating a differential index by Huffman decoding the side information bitstream; Dequantizing the differential index with a quantization table to restore additional information; Predicting a multi-channel signal using the downmix signal and the side information; Generating an upmix signal by upmixing the downmix signal based on the multi-channel signal; Extracting a time domain signal by performing an integrated filter bank on the upmix signal; And Windowing the upmix signal to extract an output signal A decoding method of a multi-channel audio based sound source position clues, characterized in that it comprises a. 9. The method of claim 8, Generating the upmix signal, Separating amplitude information and phase information from the downmix signal; Weighting the phase information by windowing a random sequence previously stored based on the amplitude information; Predicting a multi-channel signal based on the amplitude information and weighted phase information; And Upmixing the downmix signal based on the multi-channel signal A multi-channel audio decoding method based on sound source position clues, characterized in that the. 10. The method of claim 9, The separating step, Complex converting the downmix signal; And Separating amplitude information and phase information from the complex-converted downmix signal Sound channel position clue based multi-channel audio decoding method comprising a. 10. The method of claim 9, The assigning weights may include: Modeling a spectral type window for correcting phase information based on the envelope of the amplitude information; Windowing the spectral window using the pre-stored random sequence; And Weighting the phase information using a windowed random sequence Sound channel position clue based multi-channel audio decoding method comprising a. 10. The method of claim 9, Predicting the multi-channel signal, Combining the amplitude information with weighted phase information; And Predicting a multi-channel signal by inverse complex transforming the combined information in the combining step Sound channel position clue based multi-channel audio decoding method comprising a.

Images