KR100923300B1 - Method and apparatus for encoding/decoding audio data using bandwidth extension technology - Google Patents

Abstract

Disclosed are a method, an apparatus, a decoding method, and apparatus for encoding audio data using a band extension technique.

According to the present invention, there is provided a method of encoding audio data, the method comprising the steps of: (a) outputting band-limited audio data by band-extending encoding of the audio data and generating band extension information; Huffman coding with a hierarchical structure having at least one higher layer; And (c) multiplexing the Huffman coded band-limited audio data and the band extension information. As a result, the bit rate can be adjusted according to network conditions, and better quality can be ensured even when only a part of the bitstream is recovered by the decoding end.

Description

TECHNICAL FIELD AND APPARATUS FOR ENCODED / DECODED audio data using bandwidth extension technology

1 is a block diagram of an encoding apparatus according to the present invention;

2 is a detailed block diagram of the encoding apparatus of FIG. 1;

3 is a block diagram of a decoding apparatus according to the present invention;

4 is a detailed block diagram of the decoding apparatus of FIG. 3;

5 is a structural diagram of a bitstream output from the FGS encoder 2;

6 is a detailed structural diagram of additional information of FIG. 5;

7 is a structural diagram of a bitstream output from the multiplexer 3 or input to the demultiplexer 7;

8 is a reference diagram for explaining a Huffman encoding / decoding method performed in each of an encoding apparatus and a decoding apparatus according to the present invention;

9 is a reference diagram for explaining in more detail the band extension decoding performed by the BWE decoder 9, that is, BWE decoding;

10 is a flowchart for explaining an encoding method according to the present invention;

11 is a flowchart for explaining a decoding method according to the present invention.

The present invention relates to encoding and decoding of audio data, and more particularly, to an encoding method, an apparatus, a decoding method and an apparatus of audio data using a band extension technique.

With the recent development of digital signal processing technology, audio signals are mostly stored and reproduced as digital data. Digital audio storage / playback equipment samples and quantizes analog audio signals, converts them to digital signal pulse code modulation (PCM) audio data, stores them on information storage media such as CDs and DVDs, and then plays them back when needed. Allows you to listen. Digital storage / restore method of audio signal greatly improves sound quality compared to analog storage / restore methods such as LP (Long-Play Record) and magnetic tape, and significantly reduces deterioration due to the storage period. There was a problem that the storage and transmission is not so small.

In order to solve this problem, various compression schemes have been used to reduce the size of digital audio signals. AC-2 / AC-3, developed by Moving Pictures Expert Group (MPEG) / audio or Dolby, which has been standardized by the ISO (International Standard Organization), uses the human psychoacoustic model to A method of reducing the amount was adopted, and as a result, the amount of data could be efficiently reduced regardless of the signal characteristics. In other words, the MPEG / audio standard or the AC-2 / AC-3 system provides sound quality almost identical to that of a CD with only 64 Kbps-384 Kbps bit rate, which is reduced by 1/6 to 1/8 compared to previous digital coding methods. do.

However, all of these methods follow the quantization process and the encoding process to find the optimal state for a fixed bit rate. There is a problem that can not provide the above services. In addition, when a bitstream of a smaller size is converted to be suitable for a mobile device having limited storage capacity, a large amount of computation is required because a recoding process is required to reduce the size.

Accordingly, the present applicant has filed an audio encoding / decoding method and apparatus for bit rate control using Bit-Sliced Arithmetic Coding (BSAC) as Korean Patent Application No. 97-61298 filed on November 19, 1997. On April 17, 2000, it was registered as a registered patent 262653. According to BSAC, high bit rate coded bitstreams can be made into low bit rate bitstreams, and even some of the bitstreams can be recovered, resulting in network overload, poor decoder performance, or If a low bit rate is required, even if only a portion of the bitstream is degraded as the bit rate is lowered, the service can be provided to a user with a certain sound quality. Nevertheless, when the bit rate is lowered, there is a problem that performance degradation is inevitable.

In addition, BSAC has a disadvantage in that the complexity is high due to the adoption of arithmetic coding, which increases the cost when actually implementing the device. In addition, BSAC has a problem in that sound quality is worsened at a lower layer by using Modified Discrete Cosine Transform (MDCT) in converting an audio signal.

Accordingly, the present invention has been made in an effort to provide a bit rate adjustable audio encoding method, an apparatus, a decoding method, and an apparatus capable of guaranteeing good quality even if only a part of a bitstream is recovered.

Another technical problem to be achieved by the present invention is to provide a bit rate adjustable audio encoding method, a device, a decoding method and a device having a lower complexity.

Another technical problem to be solved by the present invention is to provide a bit rate adjustable audio encoding method, a decoding method, an encoding apparatus and a decoding apparatus, which can provide better sound quality even at a lower layer.

According to an aspect of the present invention, there is provided a method of encoding audio data, the method comprising the steps of: (a) outputting band-limited audio data and generating band extension information by band-extending encoding of the audio data; Encoding a Huffman into a hierarchical structure having a base layer and at least one upper layer so that bit rate is adjustable; And (c) multiplexing the Huffman coded band-limited audio data and the band extension information.

Step (b) may include: (b11) differentially encoding side information corresponding to the base layer; (b12) bit division encoding a plurality of quantized samples corresponding to the base layer; And (b13) repeating steps (b11) and (b12) for the next higher layer until encoding of a plurality of predetermined layers is completed.

Step (b) includes: (b21) differentially encoding side information including scale factor information and coding model information corresponding to the base layer; (b22) bit-dividing encoding the plurality of quantization samples corresponding to the base layer with reference to the coding model information; And (b23) repeating steps (b21) and (b22) for the next higher layer until encoding of a plurality of predetermined layers is completed.

The quantized sample is preferably one obtained by PWT conversion.

In step (c), the data corresponding to the base layer of the encoded band-limited audio data is arranged first, followed by the band extension information, and then multiplexing in the order of data corresponding to the remaining higher layers. Or the step of multiplexing in the order that the band extension information is placed first, followed by data corresponding to the base layer among the coded band-limited audio data, and then data corresponding to the remaining higher layers. Do.

Meanwhile, according to another aspect of the present invention, the technical problem is a method for decoding audio data, comprising: (a) demultiplexing an input audio bitstream and encoding a band having a base layer and a hierarchical structure having at least one upper layer Extracting limited audio data and band extension information; (b) Huffman decoding the band limited audio data corresponding to at least base layer; And (c) generating audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and adding the decoded audio data to the decoded audio data. It is also achieved by a decoding method.

Step (c) preferably includes the step of generating the audio data of the partial band to fit the boundary of the decoded audio data, the step to fit the boundary to the filter bank used in the wavelet transform Generating audio data of some bands, or interpolating a portion where the decoded audio data and the audio data of the generated some bands overlap when the filter bank used in the wavelet transform does not fit a boundary; It is more preferred to include the step of interpolation.

In the step (a), first extracting data corresponding to the base layer from the bitstream, and then extracting the bandwidth extension information, and then demultiplexing in order of extracting data corresponding to the remaining upper layers. Is preferred.

The step (a) is preferably a step of demultiplexing in order of first extracting the band extension information from the bitstream, then extracting data corresponding to the base layer, and then extracting data corresponding to the remaining upper layers. Do.

Step (b) may include: (b11) differentially decoding side information corresponding to the base layer; (b12) bit division decoding a plurality of quantized samples corresponding to the base layer; And (b13) repeating steps (b11) and (b12) for the next higher layer until decoding of a plurality of predetermined layers is completed.

Step (b) may include: (b21) differentially decoding side information including scale factor information and coding model information corresponding to the base layer; (b22) bit-decoding and decoding a plurality of quantization samples corresponding to the base layer with reference to the coding model information; And (b23) repeating steps (b21) and (b22) for the next higher layer until decoding of a plurality of predetermined layers is completed.

According to another aspect of the present invention, there is provided an apparatus for encoding audio data, the apparatus comprising: a BWE encoder outputting band-limited audio data by band-extending encoding the audio data and generating band extension information; An FGS encoder for Huffman coding the band-limited data into a hierarchical structure having a base layer and at least one upper layer to enable bit rate adjustment; And a multiplexer for multiplexing the coded band-limited audio data and the band extension information.

The FGS encoder differentially encodes additional information corresponding to the base layer, bit-divide-codes a plurality of quantization samples corresponding to the base layer, and corresponds to a next higher layer until encoding of a plurality of predetermined layers is completed. It is preferable to perform bit division coding on the side information and the plurality of quantized samples.

The FGS encoder differentially encodes additional information including scale factor information and coding model information corresponding to the base layer, and performs bit division encoding on a plurality of quantized samples corresponding to the base layer with reference to the coding model information, Encoding additional information including scale factor information and coding model information corresponding to a next higher layer and performing bit split encoding on a plurality of quantization samples corresponding to a next higher layer until encoding of a plurality of predetermined layers is completed is performed. desirable.

Preferably, the FGS encoder performs PWT conversion to obtain the quantized sample.

Preferably, the multiplexer multiplexes the data corresponding to the base layer among the encoded band limited audio data first, followed by the band extension information, and then the data corresponding to the remaining higher layers. .

Meanwhile, according to another aspect of the present invention, in the apparatus for decoding audio data, the band-limited audio data encoded in a hierarchical structure having a base layer and at least one upper layer by demultiplexing an input audio bitstream. A demultiplexer for extracting band extension information; An FGS Huffman decoder that decodes the band limited audio data corresponding to at least a base layer; And a BWE decoder generating audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and appending the decoded audio data to the decoded audio data. It is also achieved by the decoding device.

The FGS decoder differentially decodes side information corresponding to the base layer, bit-decodes a plurality of quantization samples corresponding to the base layer, and performs decoding on a next higher layer until decoding of a plurality of predetermined layers is completed. It is preferable to decode the corresponding side information and bit-decode the corresponding plurality of quantized samples.

The demultiplexer first extracts data corresponding to the base layer from the bitstream, and then extracts the band extension information, and then demultiplexes in order of extracting data corresponding to the remaining higher layers, or It is preferable to demultiplex in order of first extracting the band extension information from a stream, then extracting data corresponding to the base layer, and then extracting data corresponding to the remaining upper layers.

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

1 is a block diagram of an encoding apparatus according to the present invention.

Referring to FIG. 1, the encoding device is a device that receives Pulse Coded Modulation (PCM) audio data and encodes the same according to the present invention to output an audio bitstream. The encoding device includes a BWE encoder 1, an FGS encoder 2, and And a multiplexer 3.

The BWE encoder 1 performs band extension encoding on the PCM audio data to output band limit data and to generate band extension information. The band extension encoding refers to generating additional information necessary for recovering the data of the high frequency band which has been cut off while receiving the audio data and cutting out the data of the high frequency band over a predetermined frequency. Here, the remaining data after cutting out the data of the high frequency band among the input audio data is called band-limited audio data, and the additional information necessary for recovering the discarded data is called band extension information. A representative example of the band extension technology is Coding Technology's SBR (Spectral Band Replication) technology. A detailed description of the SBR is disclosed in Convention Paper 5560, presented at the Audio Engineering Society 112th Convention, May 10-13, 2002.

The FGS encoder 2 encodes the band limited audio data into a hierarchical structure having a base layer and at least one upper layer so that the bit rate is adjustable. FGS encoding means encoding in a plurality of hierarchical structures so that bit rate is adjustable, that is, to provide fine grain scalability (GFS). An example of FGS encoding is disclosed in Korean Patent Application No. 97-61298 filed on November 19, 1997, filed by the present applicant and disclosed in an audio encoding / decoding method and apparatus capable of adjusting bit rate on April 17, 2000. Bit split coding, that is, bit-sliced Arithmetic Coding (BSAC) coding. That is, the FGS encoder 2 differentially encodes additional information corresponding to the base layer, bit-divids-codes a plurality of quantized samples corresponding to the base layer, and then encodes the next higher layer until encoding of the plurality of predetermined layers is completed. The differential information of the layer is differentially encoded, and the corresponding plurality of quantized samples are bit-divided encoded. The additional information includes scale factor information and coding model information. Quantization samples are obtained by converting and quantizing input audio data. A more detailed description will be described later.

The multiplexer 3 multiplexes the band limited audio data encoded by the FGS encoder 2 and the band extension information generated by the BWE encoder 1.

FIG. 2 is a detailed block diagram of the encoding apparatus of FIG. 1.

Referring to FIG. 2, the encoding apparatus includes a BWE encoder 1, an FGS encoder 2, and a multiplexer 3. The same reference numerals are assigned to blocks that perform substantially the same functions as those in FIG. 1, and redundant descriptions are omitted.

In particular, the FGS encoder 2 includes a PWT converter 21, a psychoacoustic unit 22, a quantization unit 23, and an FGS Huffman encoder 24.                     

The PWT converter 21 receives PCM audio data, which is an audio signal in the time domain, and converts a PWT (Pseudo Wavelet Transform) into a signal in the frequency domain by referring to the information about the psychoacoustic model provided from the psychoacoustic unit 22. do. In the time domain, the difference in the characteristics of the audio signal perceived by human beings is not so large, but the audio signal in the frequency domain obtained through the conversion is a signal that humans can and cannot feel in each frequency band according to the human psychoacoustic model. Because of the large difference in the characteristics of, the efficiency of compression can be improved by varying the number of bits allocated to each frequency band. Compared to MDCT, where the frequency resolution in the low frequency band is higher than necessary and the deterioration perceived by the human ear is caused by small distortion, PWT conversion has a lower frequency band because the time / frequency resolution is more appropriate. Even lower layers can provide more stable sound quality.

The psychoacoustic unit 22 provides attack detection information, such as information about an acoustic psychological model, to the converting unit 21, while the audio signal converted by the converting unit 21 receives signals of appropriate subbands. By using the masking phenomenon generated by the interaction of each signal to calculate the masking threshold (masking threshold) in each subband to provide to the quantization unit (23). Masking threshold refers to the maximum size of a signal that humans do not feel due to the interaction of audio signals. In the present embodiment, the psychoacoustic unit 22 calculates masking thresholds and the like for stereo components using BMLD (Binaural Masking Level Depression).

The quantization unit 23 quantizes scalar quantization based on corresponding scale factor information of audio signals of each band such that the magnitude of quantization noise of each band is smaller than a masking threshold provided by the psychoacoustic unit 22 so that a human cannot feel it. Output samples. That is, the quantization unit 23 uses the masking threshold calculated by the psychoacoustic unit 22 and the noise-to-mask ratio (NMR), which is a ratio of noise generated in each band, to increase the NMR value of the entire band. Quantize it to 0 dB or less. An NMR value of 0 dB or less means that humans cannot hear quantization noise.

The FGS Huffman encoder 24 encodes the quantized samples and the additional information belonging to each layer and encodes them in a hierarchical structure. The additional information includes scale band information, coding band information, its scale factor information, and coding model information corresponding to each layer. The scale band information and the coding band information may be packed as header information of each frame constituting the audio bitstream and transmitted to the decoding apparatus, or may be encoded and packed as additional information for each layer and transmitted to the decoding apparatus. It may not be transmitted because it is stored in the device in advance.

More specifically, the FGS Huffman encoder 24 differentially encodes additional information including scale factor information and coding model information corresponding to the first layer, and encodes quantization samples corresponding to the first layer to corresponding coding model information. Bit division coding is performed with reference. Bit split coding is a coding scheme employed in the aforementioned BSAC encoding, which means lossless coding in order of most significant bits, next higher bits, ..., and least significant bits of quantized samples. Next, the same process is repeated for the second layer. That is, the encoding is performed while increasing the layers until the encoding of the plurality of predetermined layers is completed. The first layer is called the base layer and the rest are called higher layers. A more detailed description of the hierarchical structure will be given later.

The scale band information is information for more appropriate quantization according to the frequency characteristics of the audio signal. The scale band information indicates a scale band corresponding to each layer when the frequency domain is divided into a plurality of bands and an appropriate scale factor is allocated to each band. Say information. Thus, each layer belongs to at least one scale band. Each scale band has one scale factor assigned to it. Coding band information is also information for more appropriately performing encoding according to the frequency characteristics of an audio signal, and indicates a coding band corresponding to each layer when a frequency domain is divided into a plurality of bands and an appropriate coding model is assigned to each band. Say information. The scale band and coding band are appropriately divided by experiment, and the corresponding scale factor and coding model are determined.

The multiplexer 3 first places the data corresponding to the base layer among the encoded quantization samples, and then places the band extension information, and then places the data corresponding to the remaining higher layers or first places the band extension information. Next, the data corresponding to the base layer is placed, and then the multiplexing data is placed in the order of placing the data corresponding to the remaining upper layers.

3 is a block diagram of a decoding apparatus according to the present invention.

Referring to FIG. 3, a decoding apparatus receives an audio bitstream and decodes it according to the present invention, and outputs audio data. The decoding apparatus includes a demultiplexer 7, an FGS decoder 8, and a BWE decoder 9. do.                     

The demultiplexer 7 demultiplexes the input audio bitstream to extract band-limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer. Here, the band limited audio data and the band extension information have the same meanings as those described with reference to FIG. 1. The FGS decoder 8 decodes the band limited audio data corresponding to at least the base layer among the band limited data extracted by the demultiplexer 7. Which layer is to be deciphered depends on the network condition, the user's choice, and the like.

The BWE decoder 9 covers the data decoded by the FGS decoder 8 with reference to the band extension information extracted by the demultiplexer 7 based on the audio data decoded by the FGS decoder 8. Audio data of at least some bands which are not used is generated and added to the band limited audio data decoded by the FGS decoder 8.

On the other hand, since the present invention is based on the pseudo wavelet transform, the BWE decoder 9 goes through the following process. When encoding through Pseudo Wavelet transform, the cutoff frequency is selected by determining the last node on the frequency axis in determining band-limited audio data. Unlike MDCT, wavelet transform has low frequency resolution in high frequency part, so it is impossible to make fine adjustment when band limiting according to the determined last node. Therefore, in the decoding process, the BWE decoder 8 aligns the core portion generated by the FGS decoder 9 on the frequency axis to check the frequency bandwidth of the core portion generated by the FGS decoder 9. The BWE portion is modified and decoded accordingly.                     

For example, when restoring using only eight layers of a 64 kbps coded bit stream consisting of 16 layers, the frequency corresponding to the eighth layer is 8.5 kHz. In this case, the BWE decoder 8 restores data from 8.5 kHz to 15 kHz or more. The BWE decoder 8 can adjust the frequency bandwidth only in the bandwidth unit of one channel of QMF due to the characteristics of a quadrature mirror filter (QMF) filter. Assume that the frequency bandwidth of the nth filter of the QMF filter is 8.3 kHz. In such a case, the frequency components corresponding to 8.3 to 8.5 kHz exist on both the core part and the BWE part, so the two data must be processed properly.

The first method to deal with this is to remove all the frequency components corresponding to 8.3 ~ 8.5kHz in the core part. In this case, the FGS decoder 9 performs decoding in consideration of the bandwidth information of the BWE part, and the second method interpolates the data of the core part after the QMF filter used in the BWE decoder 8. QMF data is created by using reverse QMF filtering and restoring.

In this way, even when the audio data decoded by the FGS decoder 8 is only audio data belonging to the base band, the quality of the decoded audio data by generating and adding the audio data of the missing band by the BWE decoder 9 is added. To increase.

4 is a detailed block diagram of the decoding apparatus of FIG. 3.

Referring to FIG. 4, the decoding apparatus includes a demultiplexer 7, an FGS decoder 8, and a BWE decoder 9. Blocks that perform substantially the same function in view of the present invention are given the same reference numerals as those in FIG. 3 and duplicated descriptions are omitted.

In particular, the FGS decoder 8 is a device capable of adjusting the bit rate by decoding up to a target layer determined according to network conditions, device performance, user selection, and the like. The FGS Huffman decoder 81, the dequantizer 82, and the PWT An inverse transform unit 83 is provided. The FGS Huffman decoder 81 decodes the audio bitstream to the target layer. More specifically, Huffman-decoded coded quantization samples belonging to each layer are obtained based on coding model information obtained by decoding side information including scale factor information and coding model information corresponding to each layer to obtain quantization samples. A more detailed description will be described later.

Meanwhile, scale band information and coding band information may be obtained from header information of a bitstream or by decoding additional information of each layer. Alternatively, the decoding apparatus may previously store scale band information and coding band information.

The inverse quantization unit 82 dequantizes and restores quantized samples of each layer according to corresponding scale factor information. The PWT inverse transform unit 83 performs frequency / time mapping on the reconstructed samples and inversely converts the PWT into PCM audio data in the time domain.

The BWE decoder 9 includes a converter 91, a high frequency generator 92, an adjuster 93, and a combiner 94. The converter 91 converts the PCM audio data output from the inverse converter 83 into data in the frequency domain. The converted data is called the low frequency part. The high frequency generator 92 does not cover the data converted by the converter 91 in a manner of copying and patching a low frequency part converted by the converter 91 with reference to BWE information, that is, Create a high frequency part. The adjusting unit 93 adjusts the level of the high frequency part of the high frequency generating unit 92 by using the envelope information which is one of the BWE information. Envelope information is information sent from an encoding end and means envelope information of audio data corresponding to a high frequency part cut out by BWE encoding by an encoding end. The combining unit 94 synthesizes the low frequency portion output from the converting unit 91 and the high frequency portion output from the adjusting unit 93 to output PCM audio data.

As such, even if the FGS decoder 8 decodes only the baseband audio data, the BWE decoder 9 can improve the quality of the audio data by reconstructing and adding the audio data of the missing band.

5 shows the structure of the bitstream output from the FGS encoder 2.

Referring to FIG. 5, a frame of a bitstream encoded by the FGS encoder 2 is encoded by mapping quantization samples and additional information to a hierarchical structure for fine grain scalability (FGS). That is, it has a hierarchical structure in which the bitstream of the lower layer is included in the bitstream of the upper layer. The additional information required for each layer is divided into layers and encoded.

A header area in which header information is stored is provided at the head of the bitstream, information of layer 0 is packed, and information belonging to layers 1 to N, which is an enhancement layer, is packed in order. The header region to layer 0 information is called a base layer, and the header region to layer 1 information is called layer 1 and layer 2 information is called layer 2. In the same way, the top layer refers to the header region to layer N information, that is, from the base layer to the upper layer N. Each layer information stores additional information and coded data. For example, side information 2 and encoded quantized samples are stored as layer 2 information. Where N is an integer greater than or equal to one.

FIG. 6 shows a detailed structure of additional information of FIG. 5.

Referring to FIG. 6, additional information and coded quantization samples are stored as arbitrary layer information. In the present embodiment, since Huffman coding is performed on the quantization samples, Huffman coding model information, quantization factor information, Include additional information about the channel and other additional information. Huffman coding model information refers to index information for a Huffman coding model to be used for encoding or decoding of quantization samples belonging to a corresponding layer. The quantization factor information informs the quantization step size for quantizing or inverse magnetizing audio data belonging to the corresponding layer. The additional information about the channel refers to information about a channel such as M / S stereo. Other additional information refers to flag information on whether or not to employ the M / S stereo.

7 shows the structure of the bitstream output from the multiplexer 3 or input to the demultiplexer 7.

Referring to Fig. 7, in front of the bitstream, layer 0, which is a base layer coded by the FGS encoder 2, is placed first, followed by BWE information, followed by higher layers, that is, layer 1, layer 2, ..., layer. N is placed in order. Accordingly, even when only the base layer is received or only the base layer is decoded, the decoder can generate audio data of the missing layer based on the audio data of the base layer decoded with reference to the BWE information.

8 is a reference diagram for explaining a Huffman encoding / decoding method performed in the encoding apparatus and the decoding apparatus, respectively.

Referring to FIG. 8, the entire quantized sample to be encoded is composed of three layers. The hatched squares represent the spectral lines composed of quantized samples, the solid line represents the scale band, and the band line represents the coding band. Scale bands ①, ②, ③, ④, and ⑤ belong to layer 0, coding bands ①, ②, ③, ④, and ⑤ belong, and layer bands belong to scale bands ⑤, ⑥, and coding bands ⑥, ⑦, and ⑧. , ⑨ and 속 belong to the layer 2, and the scale bands ⑥ and ⑦ belong to the layer 2, and the coding bands ⑪, ⑫, ⑬, ⑭ and 속 belong. On the other hand, layer 0 is fixed to encode up to frequency band ⓐ, layer 1 is fixed to encode up to frequency band ⓑ, and layer 2 is fixed to encode up to frequency band ⓒ.

First, quantization samples corresponding to layer 0 within 100 bits are encoded using a coding model defined in corresponding coding bands ①, ②, ③, ④, and ⑤. As the additional information of the layer 0, the scale bands 1, 2, 3, 4 and 5 and the coding bands 1, 2, 3, 4 and 5 belonging to the layer 0 are encoded. By encoding the samples of the layer 0 in symbol units, the number of bits is counted, and when the allowed bit range is exceeded, that is, 100 bits, the encoding of the layer 0 is stopped and the layer 1 is encoded. The uncoded layer 0 samples are encoded when there is room in the bit range allowed for layer 1 and layer 2.                     

Next, quantization samples belonging to layer 1 are encoded using a coding model of a coding band belonging to layer 1, that is, a coding band to which coding quantization samples to be coded among coding bands ⑥, ⑦, ⑧, ⑨, and 이 belong. As the additional information of the layer 1, the scale bands ⑤ and ⑥ belonging to the layer 1 and the coding bands ⑥, ⑦, ⑧, ⑨ and 및 are encoded. If all the samples corresponding to layer 1 are encoded, even if the allowed bit range, that is, not 100 bits, the samples that have not been encoded in layer 0 are encoded until 100 bits are filled. When the samples corresponding to layer 1 are encoded in symbol units, the number of bits is counted, and when the allowed bit range is exceeded, that is, 100 bits, the encoding of layer 1 is stopped and the layer 2 is skipped.

Finally, the quantization samples belonging to layer 2 are encoded using a coding model of the coding band belonging to layer 2, that is, the coding band to which the quantization samples to be coded among the coding bands ⑪, ⑫, ⑬, ⑭, and 이 belong. As the additional information of the layer 2, the scale bands 6 and 7 belonging to the layer 2 and the coding bands u, u, u, u and u are encoded. If all the samples corresponding to layer 1 are encoded, even if the allowed bit range is not 100 bits, the samples that are not encoded in layer 0 are encoded until 100 bits are filled.

If all of the corresponding quantized samples are encoded without considering the allowed range of bits in layer 0 or layer 1, that is, if all of the encoded bits are encoded even though they have already exceeded the allowed range of bits, that is, 100 bits, eventually At least a part of the bit range allowed in the next layer, Layer 1, is borrowed, and it becomes impossible to encode quantized samples belonging to Layer 1. Therefore, in the case of scalable decoding, if only decoding is performed up to layer 1, the decoded quantized samples may rise or fall below the frequency ⓑ because the decoding is not performed until the frequency ⓑ of the layer 1. At this time, a birdy effect appears that deteriorates the sound quality.

On the other hand, when determining a plurality of layers (target layers), a bit range is allocated in consideration of the size of the entire audio data to be encoded, so that the encoding cannot be generated because the bit range to be encoded is insufficient.

Like the encoding, the decoding process also counts the number of bits according to the allowable bit range while performing the reverse process, so that it is possible to find a time to decode to layer 1.

9 is a reference diagram for explaining in more detail the band extension decoding performed by the BWE decoder 9, that is, BWE decoding.

Referring to Fig. 9, the striped portion represents data decoded by the FGS decoder 8, and the gray portion represents data generated by the BWE decoder 9. When data up to one-quarter of the sampling frequency Fs belongs to the base layer, (a) indicates that only the data corresponding to the base band is decoded by the decoding stage, and (b), (c), and (d) The case where data belonging to the base layer and at least one upper layer is decoded by the FGS decoder 8 is shown. That is, the FGS decoder 8 can decode the data so that the bit rate can be adjusted, and the BWE decoder 9 generates data of missing bands that the FGS decoder 8 cannot decode.

A coding method and a decoding method according to a preferred embodiment of the present invention will be described below based on the above configuration.

10 is a flowchart for explaining an encoding method according to the present invention.

Referring to FIG. 10, the encoding apparatus band-extends audio data to output band-limited audio data and generates band extension information for a base layer (step 1001). The meaning of the band extension information for the base layer is information for the decoder to generate audio data of the remaining missing band based on the audio data belonging to the base layer, and includes envelope information and the like. Next, the encoding apparatus encodes the band-limited data into a hierarchical structure having a base layer and at least one upper layer so that the bit rate can be adjusted. More specifically, band-limited audio data is PWT-converted for each layer (step 1002), quantized (step 1003), Huffman coded, and packaged in a hierarchical structure to enable bit rate adjustment (step 1004). Finally, the audio bitstream obtained by multiplexing the encoded band-limited audio data and the band extension information is output (step 1004). More specifically, the encoding apparatus multiplexes or bands in a sequence in which data corresponding to a base layer among the encoded band limited audio data is placed first, followed by band extension information, and then data corresponding to the remaining upper layers are arranged. The extension information is placed first, followed by data corresponding to the base layer, and then multiplexed in the order of data corresponding to the remaining upper layers.                     

11 is a flowchart for explaining a decoding method according to the present invention.

Referring to FIG. 11, the decoding apparatus demultiplexes an input audio bitstream to extract band-limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer (step 1101). That is, the data corresponding to the base layer is first extracted from the input audio bitstream, and then the band extension information is extracted, followed by demultiplexing in the order of extracting the data corresponding to the remaining upper layers, or the first band extension information. We then extract the data corresponding to the base layer, and then demultiplex in the order of extracting the data corresponding to the remaining upper layers. Subsequently, the decoding apparatus decodes the band-limited audio data corresponding to at least the base layer to be bit rate adjustable. More specifically, Huffman decoding to the target layer (step 1102), inverse quantization (step 1103), and PWT inverse transformation (step 1104) to obtain band-limited PCM audio data. Next, the PCM audio data of at least a part of the band not covered by the audio data obtained in step 1104 is generated based on the PCM audio data obtained in step 1104, and is output in addition to the PCM audio data obtained in step 1104. (Step 1105).

As described above, according to the present invention, there is provided a bit rate adjustable audio encoding method, an apparatus, a decoding method, and an apparatus capable of guaranteeing better quality even if only a part of a bitstream is recovered.

It also has lower complexity and better sound quality even at lower levels. Compared to MPEG-4 Audio BSAC using arithmetic coding, the encoding / decoding apparatus of the present invention using Huffman coding greatly reduces the amount of computation during bit packing / unpacking. Even if bit packing according to the present invention is performed to provide the FGS, the overhead is small, and is similar to the case where scalability is not provided in terms of encoding gain.

In addition, it is possible to provide a seamless service by changing the transmission bit rate according to the user's will or network environment when transmitting the audio stream through the network. When storing in an information storage medium having a limited capacity, the file size can be arbitrarily adjusted and stored. When the bit rate is lowered, the bandwidth is limited, and thus the complexity of the filter, which mainly accounts for most of the complexity of the encoding / decoding device, is considerably reduced, so that the actual complexity of the encoding device / decoding device is inversely proportional to the bit rate.

In addition, by adopting the PWT transform, the resolution of the time / frequency axis is superior to that of the conventional MDCT-based encoding, thereby providing better sound quality at a lower layer.

Claims (23)

delete In a method of encoding audio data, (a) band extending encoding the audio data to output band limited audio data and generating band extension information; (b) Huffman encoding the band limited data into a hierarchical structure having a base layer and at least one upper layer to enable bit rate control; And (c) multiplexing the Huffman coded band-limited audio data and the band extension information, Step (b) is (b11) differentially encoding side information corresponding to the base layer; (b12) bit division encoding a plurality of quantized samples corresponding to the base layer; And and (b13) repeating steps (b11) and (b12) for the next higher layer until encoding for a plurality of predetermined layers is completed. In a method of encoding audio data, (a) band extending encoding the audio data to output band limited audio data and generating band extension information; (b) Huffman encoding the band limited data into a hierarchical structure having a base layer and at least one upper layer to enable bit rate control; And (c) multiplexing the Huffman coded band-limited audio data and the band extension information, Step (b) is (b21) differentially encoding side information including scale factor information and coding model information corresponding to the base layer; (b22) bit-dividing encoding the plurality of quantization samples corresponding to the base layer with reference to the coding model information; And and (b23) repeating steps (b21) and (b22) for the next higher layer until encoding of a plurality of predetermined layers is completed. The method according to claim 2 or 3, And the quantized sample is obtained by PWT conversion. The method according to claim 2 or 3, Step (c) is And encoding the data corresponding to the base layer among the encoded band-limited audio data first, followed by the band extension information, and then multiplexing the data corresponding to the remaining higher layers. Way. The method according to claim 2 or 3, Step (c) is Wherein the band extension information is arranged first, followed by multiplexing in the order that data corresponding to the base layer among the encoded band-limited audio data is arranged, and then data corresponding to the remaining upper layers are arranged. Way. delete In the method of decoding audio data, (a) demultiplexing the input audio bitstream to extract band limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer; (b) Huffman decoding the band limited audio data corresponding to at least base layer; And (c) generating audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and adding the decoded audio data to the decoded audio data; Step (c) is And generating the audio data of the partial band so as to conform to the boundary of the decoded audio data. The method of claim 8, Step (c) is And generating the audio data of the partial band so as to fit the boundary of the filter bank used in the wavelet transform. The method of claim 8, Step (c) is Interpolating the overlapped portion of the decoded audio data and the generated partial band audio data when the filter bank used in the wavelet transform does not fit the boundary. Decryption method The method of claim 8, Step (a) is First extracting data corresponding to the base layer from the bitstream, and then extracting the band extension information, and then demultiplexing in the order of extracting data corresponding to the remaining upper layers. . The method of claim 8, Step (a) is And demultiplexing in the order of first extracting the band extension information from the bitstream, then extracting data corresponding to the base layer, and then extracting data corresponding to the remaining upper layers. In the method of decoding audio data, (a) demultiplexing the input audio bitstream to extract band limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer; (b) Huffman decoding the band limited audio data corresponding to at least base layer; And (c) generating audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and adding the decoded audio data to the decoded audio data; Step (b) is (b11) differentially decoding side information corresponding to the base layer; (b12) bit division decoding a plurality of quantized samples corresponding to the base layer; And and (b13) repeating steps (b11) and (b12) for the next higher layer until decoding of a plurality of predetermined layers is completed. In the method of decoding audio data, (a) demultiplexing the input audio bitstream to extract band limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer; (b) Huffman decoding the band limited audio data corresponding to at least base layer; And (c) generating audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and adding the decoded audio data to the decoded audio data; Step (b) is (b21) differentially decoding side information including scale factor information and coding model information corresponding to the base layer; (b22) bit-decoding and decoding a plurality of quantization samples corresponding to the base layer with reference to the coding model information; And (b23) repeating steps (b21) and (b22) for the next higher layer until decoding of a plurality of predetermined layers is completed. delete In the apparatus for encoding audio data, A BWE encoder for performing band extension encoding on the audio data to output band limited audio data and to generate band extension information; An FGS encoder for Huffman coding the band-limited data into a hierarchical structure having a base layer and at least one upper layer to enable bit rate adjustment; And A multiplexer for multiplexing the coded band-limited audio data and the band extension information, The FGS coder Differentially encoding the additional information corresponding to the base layer, bit-dividing-code the plurality of quantized samples corresponding to the base layer, and additional information corresponding to the next higher layer until encoding of the plurality of predetermined layers is completed; An encoding device characterized by performing bit division encoding on a plurality of quantized samples. In the apparatus for encoding audio data, A BWE encoder for performing band extension encoding on the audio data to output band limited audio data and to generate band extension information; An FGS encoder for Huffman coding the band-limited data into a hierarchical structure having a base layer and at least one upper layer to enable bit rate adjustment; And A multiplexer for multiplexing the coded band-limited audio data and the band extension information, The FGS coder Differentially encode additional information including scale factor information and coding model information corresponding to the base layer, and perform bit division coding on a plurality of quantized samples corresponding to the base layer with reference to the coding model information, Encoding, wherein the additional information including scale factor information and coding model information corresponding to the next higher layer is encoded and bit-divided encoding of a plurality of quantized samples corresponding to the next higher layer until encoding on the layer is completed. Device. The method according to claim 16 or 17, The FGS coder And a quantized sample is obtained by PWT conversion. The method according to claim 16 or 17, The multiplexer And encoding the data corresponding to the base layer among the encoded band-limited audio data first, followed by the band extension information, and then multiplexing the data corresponding to the remaining higher layers. . delete In the apparatus for decoding audio data, A demultiplexer for demultiplexing the input audio bitstream to extract band-limited audio data and band extension information encoded in a hierarchical structure having a base layer and at least one upper layer; A FGS decoder for Huffman decoding the band limited audio data corresponding to at least base layer; And A BWE decoder that generates audio data of at least some bands not covered by the decoded audio data based on the decoded audio data and appends the decoded audio data to the decoded audio data; The FGS decoder Differentially decode the additional information corresponding to the base layer, bit-decode and decode a plurality of quantized samples corresponding to the base layer, and add additional information corresponding to a next higher layer until decoding of a plurality of predetermined layers is completed Decoding and bit-decoding the corresponding plurality of quantized samples. The method of claim 21, The demultiplexer is And demultiplexing the data corresponding to the base layer first from the bitstream, subsequently extracting the band extension information, and then extracting data corresponding to the remaining upper layers. The method of claim 21, The demultiplexer is And demultiplexing in order of first extracting the band extension information from the bitstream, subsequently extracting data corresponding to the base layer, and then extracting data corresponding to the remaining upper layers.

Images