KR100370412B1 - Audio decoding method with adjustable complexity and audio decoder using the same - Google Patents

Abstract

PURPOSE: An audio decoding method for controlling complexity and an audio decoder using the method are provided to control complexity according to performance of a platform in which a decoder is constructed. CONSTITUTION: An encoded mono-mode audio bit stream is analyzed according to syntax. Bit information and a scale factor allocated to each of bands divided according to the analysis step are decoded in a scale factor decoder(34). Quantized audio data of each of the divided bands is inverse-quantized using the decoded bit allocation information. The inverse-quantized audio data of each of the divided bands is multiplied by the decoded scale factor, to obtain inverse-quantized band split signals. Signals of upper bands is removed from the inverse-quantized band split signals to restrict the bands. The bands of the band split signals whose bands have been restricted are synthesized to convert the signals into audio data in time domain.

Description

Audio decoding method with adjustable complexity and audio decoder using the same

The present invention relates to an audio decoding method and an audio decoder, and more particularly, to an audio decoding method capable of adjusting the complexity and a decoder using the same.

The Moving Picture Experts Group (MPEG) audio standard defines a new way of compressing high-quality audio signals at low bit rates. Among them, MPEG-1 audio coding is 32-192 kbit / s for mono audio channels and 64 to 384 kbit / s for stereo audio channels at sampling frequencies of 32 kHz, 44.1 kHz, and 48 kHz. This is the first international standard for compressing audio signals. Subsequently, MPEG-2 audio coding has been extended to audio signal compression and multichannel audio signal compression at low sampling frequencies of 16 kHz, 22.05 kHz, and 24 kHz. The MPEG audio standard delivers sound quality nearly the same as that of a compact disc at a bit rate that is 1/6 to 1/8 times less than conventional digital coding. For this reason, the MPEG audio standard will play an important role in the storage and transmission of audio signals such as digital audio broadcasting (DAB), Internet phones, audio on demand and multimedia systems.

In the case of a conventional encoder designed for low bit rate encoding, only the case where all processes of encoding / decoding are performed is considered. That is, when implemented in the platform (platform), since only the case of always providing a certain level of sound quality, therefore, the platform that can be implemented in this way was not limited, but was bound to be limited. For example, since the implementation of the MPEG-1 audio decoder as shown in FIG. 1 requires a considerable amount of computation, research has been conducted to reduce the computational amount, but even a more optimized codec has been developed. It requires computation. Therefore, when implemented with software alone, it does not work on PCs below 486, but only on PCs above Pentium.

On the other hand, in order to be able to operate on a lower performance PC, it is necessary to use a codec structure that can reduce the complexity and reduce the performance while using the bit stream of MPEG-1 audio as it is. When the complexity can be adjusted in this way, when the decoder is implemented in hardware, the sound quality may be somewhat lower, but since a decoder can be implemented at a low price, various types of decoders can be provided according to a user's preference. That is, since the performance of the platform on which the MPEG-1 audio decoder can be implemented varies widely, if the platform performance is excellent to satisfy this diversity, the audio signal of excellent sound quality is restored through all the procedures defined in the MPEG-1 audio. can do. However, if the platform does not perform well, a modified MPEG-1 audio decoder is needed that can reconstruct a signal with slightly degraded sound quality instead of reducing complexity by analyzing the MPEG-1 audio bitstream and having only the necessary information.

Accordingly, an object of the present invention is to provide an audio decoding method that has the same bitstream structure as MPEG-1 audio and can adjust the complexity according to the performance of a platform on which a decoder is implemented, and obtain a corresponding sound quality according to the complexity. .

Another object of the present invention is to provide an audio decoder using the audio decoding method capable of adjusting the complexity.

In order to achieve the above object, an audio decoding method capable of adjusting the complexity according to the present invention

Analyzing the encoded mono bit audio bitstream according to the syntax;

Decoding bit information and scale factors allocated to each divided band separated by the analyzing step;

Dequantizing the quantized audio data of the respective divided bands using the decoded bit allocation information;

Obtaining dequantized band split signals by multiplying the decoded scale factor by the dequantized audio data of each divided band;

Limiting a band by removing some upper band signals from the dequantized band split signals; And

And synthesizing the bands of the band-limited band split signals into audio data of a time domain.

In order to achieve the above object, an audio decoding method capable of adjusting the complexity according to the present invention

Analyzing the audio bitstream of the encoded stereo mode according to the syntax;

Decoding the bit information and scale factor allocated to each divided band separated by the analyzing step with respect to the left channel, and inverse quantizing the quantized audio data of each divided band using the decoded bit allocation information, Obtaining dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the decoded scale factor;

Decoding the bit information and scale factor allocated to each divided band separated by the analyzing step for the right channel, and inverse quantizing the quantized audio data of each divided band using the decoded bit allocation information, Obtaining dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the decoded scale factor;

Adding dequantized band split signals with respect to the left channel and dequantized band split signals with respect to the right channel; And

And synthesizing the bands of the dequantized band split signals generated by the addition step into audio data in the time domain.

In order to achieve the above another object, an audio decoder capable of adjusting the complexity according to the present invention

A bitstream analyzer for analyzing the encoded monobit audio bitstream according to syntax and separating the encoded audio bitstream into pieces of information necessary for decoding;

A bit allocation information decoder for decoding the bit information allocated to each divided band separated from the bitstream analyzer;

A sample dequantizer for dequantizing the quantized audio data of each split band using the decoded bit allocation information;

A scale factor decoder for decoding a scale factor for each divided band separated from the bitstream analyzer;

A multiplier for multiplying dequantized audio data of each divided band by a scale factor of each divided band to obtain dequantized band split signals;

A band limiter for limiting a band by removing some upper band signals from the dequantized band split signals output from the multiplier; And

And a synthesis filter for synthesizing the bands of the band split signals output from the band limiter and converting the bands into audio data in the time domain.

In order to achieve the above another object, an audio decoder capable of adjusting the complexity according to the present invention

A bitstream analyzer for analyzing the encoded audio bitstream of the stereo mode according to the syntax and separating the encoded audio bitstream into pieces of information necessary for decoding;

A first bit allocation information decoder for decoding bit information allocated to each divided band for the left channel separated from the bitstream analyzer;

A first sample inverse quantizer for inversely quantizing the quantized audio data of each divided band for the left channel using the decoded bit allocation information;

A first scale factor decoder for decoding a scale factor for each division band for the left channel separated from the bitstream analyzer;

A first multiplier configured to multiply the inverse quantized audio data of each divided band by a scale factor of the respective divided bands with respect to the left channel to obtain dequantized band split signals;

A second bit allocation information decoder for decoding bit information allocated to each divided band for the right channel separated from the bitstream analyzer;

A second sample inverse quantizer for inversely quantizing the quantized audio data of each divided band for the right channel using the decoded bit allocation information;

A second scale factor decoder for decoding a scale factor for each division band for the right channel separated from the bitstream analyzer;

A second multiplier for obtaining the dequantized band split signals by multiplying the dequantized audio data of the respective divided bands by 1/2 and the scale factor of each divided band for the right channel;

An adder for adding dequantized band split signals with respect to the left channel output from the first multiplier and dequantized band split signals with respect to the right channel output from the second multiplier;

A band limiter for limiting a band by removing signals of some upper bands from the dequantized band split signals output from the adder; And

And a synthesis filter for synthesizing the bands of the band split signals output from the band limiter and converting the bands into audio data in the time domain.

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

3 is a block diagram showing a configuration of an audio decoder according to an embodiment of the present invention, which includes a bitstream analyzer 31, a bit allocation information decoder 32, a sample dequantizer 33, and a scale factor decoding. Group 34, multiplier 35, band limiter 36 and band-limited synthesis filter 37.

Referring to the operation and effects of the audio decoder according to an embodiment of the present invention shown in Figure 3 as follows.

The bitstream analyzer 31 separates each piece of information necessary for decoding according to the bitstream syntax with respect to the input coded bitstream, and allocates a bit allocation information decoder 32, a sample dequantizer 33, and a scale factor. It is supplied to the decoder 34.

The bit allocation information decoder 32 decodes the bit information allocated to each band extracted from the bitstream analyzer 31 and supplies it to the sample dequantizer 33.

The sample dequantizer 33 obtains dequantized data by using the quantized data of each band and the decoded bit allocation information, and supplies the dequantized data to the multiplier 35.

The scale factor decoder 34 decodes the scale factor for each band extracted from the bitstream analyzer 31 and supplies it to the multiplier 35.

The multiplier 35 multiplies the normalized data of each dequantized band by the scale factor of each band to obtain the actual data of each band, and supplies it to the band limiter 36.

In order to reduce the complexity of the decoder, the band limiter 36 removes some upper band signals from the dequantized band split signals to forcibly limit the band. In this case, the degree of complexity can be controlled by adjusting the degree of bandwidth limitation according to the needs of users, the performance of the system or the intention of the manufacturer.

In the band-limited synthesis filter 37, band-limited band split signals are performed as time-domain signals.

Here, before the operation of the band limiter 36 and the band-limited synthesis filter 37, which are the main components of the present invention, in more detail, the operation of the conventional synthesis filter (16 in FIG. 1) is shown in FIG. Referring to Figure 2 and as follows.

When the signals of each band are restored from the bitstream analyzer (11 in FIG. 1) through the multiplier 15, they are converted into signals in the time domain through the synthesis filter (16 in FIG. 1) represented by the following equation (1). .

In formula (1), D (n) and Is defined as the following equation (2) and (3). D (n) is a coefficient of the synthesis window, and the original prototype window f [n] is modified and used in advance in equation (2) instead of f [n] in the MPEG-1 audio standard. On the other hand, M is the number of divided bands, and 32 is used for MPEG-1 audio.

In the above formula (3) Is the signal of the l-th band of the m-th block including the quantization noise through quantization and inverse quantization. N [n] [ℓ] is a matrix calculated as in the following equation (4).

As such, the synthesis filter (16 in FIG. 1) recommends in the MPEG-1 audio standard to implement the above formulas (1) to (4) in the process shown in FIG.

Meanwhile, in the present invention, the band-divided signals are divided into a lower band from 0 to (M / 2-1) th band and an upper band from M / 2 to M-1th band, and the band limiter (see FIG. In 36), all signals in the upper band are removed to be zero, and then passed through a synthesis filter (37 in FIG. 3) to obtain a restored signal. However, aliasing components generated between neighboring bands in the analysis filter (not shown) of the encoder because the signal of the upper band is forcibly removed to limit the band before passing the synthesis filter 37. Some of these are not compensated in the synthesis filter 37. That is, since most of the aliasing components are generated between neighboring bands, the aliasing components of the band located at the boundary portion cannot be compensated. However, the degradation of sound quality due to the influence of such an uncompensated aliasing component is not so large. Since the restored signal is a band-limited signal, the characteristic of the signal does not change even when downsampling the output signal. Looking at the expression (1), even number Is even Only odd number Is odd It is only related. Therefore, even number Is even If only the value is known, it can be obtained by the above equation (1), so that the calculation amount related to the above equation (1) can be reduced by about 1/2. In this principle, the amount of computation can be reduced by band limitation.

Applying this principle, the band synthesis process of FIG. 2 may be modified and implemented as shown in FIG. If the band is limited to the 15th band, Q = 2 in FIG.

Likewise, if the band is divided into M / 4 and M / 8 bands to make the high band signal zero, the bandwidth is reduced correspondingly to restore the signal which is considerably band limited compared to the original signal, but the amount of calculation It is possible to provide a trade-off which is reduced to less than 1/4, 1/8 of. The band-limited decoder has the structure of a decoder in the case of Q = 4 and Q = 8 in FIG.

Among these decoding processes, the synthesis filter part occupies more than 90% of the total calculation amount. Therefore, the change of the total calculation amount is determined according to the calculation amount of the synthesis filter portion. In order to further optimize the synthesis filter part, that is, according to the fast filter algorithm, the filter equation of Equation (1) is rearranged as in Equation (5).

In Equation (5), z [n] is obtained from band-divided signals as in Equation (6).

Equation (6) is a general discrete cosine transform (DCT) form, so it can be implemented using a fast DCT algorithm. The fast algorithm for the DCT of Equation (6) has a structure as shown in FIG.

The fast DCT algorithm shown in FIG. 5 corresponds to the case of M = 8, and by extending this structure in the same form, a structure for M = 32 such as the MPEG-1 audio standard can be obtained. On the other hand, the method of reducing the amount of calculation due to band limitation can be applied to the case of using a fast synthesis filter. If band limit is in M / 2 band, even number Find the Bay and Odd Number Since only the hatched portion of FIG. 5 needs to be calculated, the amount of computation related to DCT is reduced by about 1/2. Similarly, as shown in FIG. 5, even when the band limit is limited to the M / 4 and M / 8th bands, the number of operations associated with the filter is reduced, and the amount of calculation is reduced to 1/4 or 1/8 or less. Therefore, filters that can reduce the complexity can be implemented using the structure of the proposed fast filter algorithm.

On the other hand, it is possible to implement a decoder having a variety of complexity depending on the needs of the system or user. When the decoder is implemented using the above-described method, various combinations of complexity and performance are possible, so that a decoder suitable for the performance of the implemented system can be provided. Also, when implemented in hardware, various decoders of various types depending on price can be provided. The ability to create devices allows you to choose between your performance and price.

Also, audio data is generally in stereo. Therefore, when encoding is performed in the stereo mode, the amount of computation can be reduced to about 1/2 of the stereo mode by decoding in the mono mode.

6 is a block diagram showing a configuration according to another embodiment of an audio decoder according to the present invention for synthesizing a stereo bitstream into a mono signal. The bitstream analyzer 61, the first and second bit allocation information decoders ( 62,66), first and second sample inverse quantizers 63,67, first and second scale factor decoders 64,68, first and second multipliers 65,69, adder 70 ) And a synthesis filter (71).

Referring to the operation and effect of the audio decoder according to another embodiment of the present invention shown in FIG.

The left channel signal restored by the equation (5) Signal from the right channel Speaking of mono signals created by these two signals Can be expressed as

And, and If obtained by the above formula (5) and substituted into the above formula (7) can be summarized as follows (8).

In the above formula (8) Is obtained by the following equation (9).

In the above formula (9) and Denotes the dequantized signal of the band-divided signal in the left channel and the right channel, respectively.

6 again, the bitstream analyzer 61 separates necessary information according to syntax from the encoded bitstreams for the left channel and the right channel, respectively.

The first bit allocation information decoder 62 and the first scale factor decoder 64 decode the bit information and scale factor allocated to each division band for the left channel separated from the bitstream analyzer 61, respectively. The first sample dequantizer 63 dequantizes the quantized audio data of each divided band for the left channel by using the decoded bit allocation information. The first multiplier 65 obtains dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the scale factor of each divided band for the left channel.

The second bit allocation information decoder 66 and the second scale factor decoder 68 decode bit information and scale factors allocated to respective division bands for the right channel separated from the bitstream analyzer 61, respectively. The second sample dequantizer 67 dequantizes the quantized audio data of each divided band for the right channel by using the decoded bit allocation information. The second multiplier 69 obtains dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the scale factor of each divided band for the left channel.

In the adder 70, the dequantized band split signals with respect to the left channel output from the first multiplier 65 and the dequantized band split signals with respect to the right channel output from the second multiplier 69, (7) The dequantized data of the left channel and the right channel can be added by the equations (9), and the added data can be restored to the mono data by passing through the synthesis filter 71. That is, by restoring the stereo data to the mono data through the process shown in FIG. 6, the overall sound quality is slightly degraded, but the computation amount can be reduced.

As described above, the audio decoding method and the decoder which can adjust the complexity according to the present invention adjust the complexity of the synthesis filter by removing the signals of some upper band from the dequantized band split signals to limit the band, Although the sound quality of the reconstructed audio is slightly lower than that of the decoder recommended in the MPEG-1 audio standard, various decoders can be implemented that can greatly reduce the amount of computation.

In addition, when restoring the encoded stereo bitstream, by restoring mono data by separately decoding the left channel and the right channel, the sound quality of the restored audio is somewhat lower than that of the decoder recommended in the MPEG-1 audio standard. Only a decoder that can significantly reduce the amount of computation can be implemented.

In addition, it is possible not only to implement a decoder suitable for the performance of each system, but when applied to a system having excellent performance, a relationship between performance and complexity can be adjusted according to a user's request without adding memory.

1 is a block diagram showing the configuration of a typical MPEG-1 audio decoder.

2 is a flowchart illustrating the operation of the band synthesis filter in FIG. 1;

3 is a block diagram showing a configuration of an audio decoder according to an embodiment of the present invention.

4 is a flowchart for explaining the operation of the band-limited synthesis filter in the (32 / Q-1) th band in FIG.

FIG. 5 is a structural diagram of a fast DCT (when M = 8) to be used for a fast band synthesis filter when the band is limited in FIG.

6 is a block diagram showing a configuration according to another embodiment of an audio decoder according to the present invention for synthesizing an encoded stereo bitstream into a mono signal.

* Explanation of symbols for main parts of the drawings

31,61 ... Bitstream Analyzer

32,62,66 ... bit allocation information decoder 33,63,67 ... sample dequantizer

34,64,68 ... scale factor decoder 35,65,69 ... multiplier

36 ... band limiter 37 ... band limited synthesis filter

70 ... adder 71 ... synthetic filter

Claims (9)

Analyzing the encoded mono bit audio bitstream according to the syntax; Decoding bit information and scale factors allocated to each divided band separated by the analyzing step; Dequantizing the quantized audio data of the respective divided bands using the decoded bit allocation information; Obtaining dequantized band split signals by multiplying the decoded scale factor by the dequantized audio data of each divided band; Limiting a band by removing some upper band signals from the dequantized band split signals; And And converting the bands of the band-limited band split signals into audio data of a time domain. The method of claim 1, wherein when the number of bands divided during encoding is M, in order to reduce the amount of calculation in the band synthesis step to 1 / Q in the MPEG-1 audio standard, the band limiting step is performed. An audio decoding method with complexity control, characterized in that the band is limited to the (M / Q-1) th band. 3. The audio decoding method of claim 2, wherein in the band synthesis step, a fast DCT structure is used. A bitstream analyzer for analyzing the encoded monobit audio bitstream according to syntax and separating the encoded audio bitstream into pieces of information necessary for decoding; A bit allocation information decoder for decoding the bit information allocated to each divided band separated from the bitstream analyzer; A sample inverse quantizer for dequantizing quantized audio data of each partition band using the decoded bit allocation information: A scale factor decoder for decoding a scale factor for each divided band separated from the bitstream analyzer; A multiplier for multiplying dequantized audio data of each divided band by a scale factor of each divided band to obtain dequantized band split signals; A band limiter for limiting a band by removing some upper band signals from the dequantized band split signals output from the multiplier; and An audio decoder having a complexity control, characterized in that it comprises a synthesis filter for synthesizing the bands of the band split signals output from the band limiter and converting them into audio data in a time domain; 5. The method of claim 4, wherein when the number of bands divided during encoding is M, in order to reduce the amount of computation in the synthesis filter to 1 / Q in the MPEG-1 audio standard, An audio decoder having a complexity control, characterized in that the band is limited to the M / Q-1) th band. 6. The audio decoder of claim 5, wherein the synthesis filter uses a fast DCT structure. Analyzing the audio bitstream of the encoded stereo mode according to the syntax; Decoding the bit information and scale factor allocated to each divided band separated by the analyzing step with respect to the left channel, and inverse quantizing the quantized audio data of each divided band using the decoded bit allocation information, Obtaining dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the decoded scale factor; Decoding the bit information and scale factor allocated to each divided band separated by the analyzing step for the right channel, and inverse quantizing the quantized audio data of each divided band using the decoded bit allocation information, Obtaining dequantized band split signals by multiplying the dequantized audio data of each divided band by 1/2 of the decoded scale factor; Adding dequantized band split signals with respect to the left channel and dequantized band split signals with respect to the right channel; and And combining the bands of the dequantized band split signals generated by the adding step into the audio data of the time domain. 8. The complexity control method of claim 7, wherein the audio decoding method further comprises limiting a band by removing a part of higher-order signals from dequantized band split signals generated by the adding step. Possible audio decoding methods. A bitstream analyzer for analyzing the encoded audio bitstream of the stereo mode according to syntax and separating the encoded audio bitstream into pieces of information necessary for decoding; A first bit allocation information decoder for decoding bit information allocated to each divided band for the left channel separated from the bitstream analyzer; A first sample inverse quantizer for inversely quantizing the quantized audio data of each divided band for the left channel using the decoded bit allocation information; A first scale factor decoder for decoding a scale factor for each division band for the left channel separated from the bitstream analyzer; A first multiplier configured to multiply the inverse quantized audio data of each divided band by a scale factor of the respective divided bands with respect to the left channel to obtain dequantized band split signals; A second bit allocation information decoder for decoding bit information allocated to each divided band for the right channel separated from the bitstream analyzer; A second sample inverse quantizer for inversely quantizing the quantized audio data of each divided band for the right channel using the decoded bit allocation information; A second scale factor decoder for decoding a scale factor for each division band for the right channel separated from the bitstream analyzer; A second multiplier configured to multiply the inverse quantized audio data of each divided band by a scale factor of the respective divided bands with respect to the right channel to obtain dequantized band split signals; An adder for adding dequantized band split signals with respect to the left channel output from the first multiplier and dequantized band split signals with respect to the right channel output from the second multiplier; A band limiter for limiting a band by removing signals of some upper bands from the dequantized band split signals output from the adder; And And a synthesis filter for synthesizing the bands of the band split signals output from the band limiter and converting the bands into audio data in a time domain.