KR100433984B1 - Method and Apparatus for Encoding/decoding of digital audio - Google Patents

Abstract

The present invention relates to a digital audio encoding / decoding apparatus and method capable of enhancing the efficiency of a storage medium and increasing transmission efficiency by improving the audio quality at the same compression rate while maintaining audio quality, The digital audio encoding apparatus includes a down-sampling unit 100 for receiving a digital audio signal and down-sampling the digital audio signal 1 / N times and outputting the result; An audio encoding unit 200 for encoding the downsampled audio signal output from the downsampling unit 100 using a digital audio encoding scheme; A frequency band that can not be encoded is modeled by using a gain of a frequency band between a spectrum of an input digital audio signal and a downsampled signal output from the audio encoding unit 200, A spectral modeling unit 300; A spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit 300 and an encoded bitstream output from the audio encoder are multiplexed to maintain compatibility with the digital audio encoding standard, And a multiplexing unit 400 for multiplexing the data.

Description

[0001] The present invention relates to a digital audio encoding / decoding apparatus and a digital audio encoding /

The present invention relates to a digital audio encoding / decoding apparatus and method, and more particularly, to a digital audio encoding / decoding apparatus and method for performing a digital audio signal to be stored in a limited storage medium or transmitted in a limited transmission band will be.

That is, the present invention relates to a technique for improving audio quality or increasing a compression ratio by solving the problems in the conventional Perceptual Audio Coding scheme.

A compact disc (CD), which is the beginning of digital audio, samples 44,100 samples per second of an analog stereo (left and right) audio signal at a 16-bit resolution, And then reads it optically to reproduce the audio.

Since 72 minutes of music is stored in one CD and its data amount is about 1.3 gigabyte (Gbyte), it is necessary to store a large amount of memory in a storage medium such as a hard disk, It was not suitable for real-time transmission in a transmission environment.

Accordingly, techniques for compressing audio and reducing storage capacity to effectively transmit or store such large-capacity digital audio data have been developed.

In the human ear, a masking effect is generated in which the surrounding signals are inaudible due to a relatively large sound on the frequency axis or the time axis, and a method of compressing the audio signal using the above- It is called a Perceptual Audio Coding method.

Under the leadership of Moving Picture Experts Group (MPEG), the perceptual audio coding method has been internationally standardized since 1992, and Dolby, Sony, ATT (MP3) (hereinafter referred to as " MP3 "), which we often hear, refers to the MPEG-1 audio layer III.

However, due to the rapid expansion of the Internet environment and the speeding up of the Internet environment, downloading and enjoying music files such as the MP3 as a home page or an email has become a routine task. In 1999, portable MP3 players It became easy to listen to digital audio anytime and anywhere.

In addition, methods such as "RTP / RTCP" for transmitting voice, audio and video data requiring real-time transmission have appeared in which the entire multimedia data is downloaded and reproduced without receiving the entire multimedia data.

However, in order to maintain audio quality of stereo (2-channel) audio similar to CD using MP3, a transmission rate of 128 kbps or more is required, and a transmission rate of 96 kbps or more is required when the MPEG-2 AAC compression algorithm is used.

Of course, in a high-speed Internet environment, a music file can be transmitted at a transmission rate of 128 kbps or more. However, a maximum transmission rate is variable depending on a change of a user's connection state or a network environment. Particularly in a mobile communication environment, Is reduced.

Accordingly, in order to transmit a music file in real time while maintaining sound quality at a proper level, a lower transmission rate should be used. In addition, in order to simultaneously transmit video and audio in a limited band, most of the band is allocated to a video having a large amount of data There is a problem that the quality of audio is deteriorated because the band for audio is inevitably shrunk.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to solve the problems of the prior art as described above and to provide a method and apparatus for enhancing the efficiency of a storage medium by increasing the compression ratio or improving the audio quality at the same compression rate while maintaining audio quality, Decoding apparatus and method capable of increasing the number of digital audio encoding / decoding apparatuses and methods.

Another object of the present invention is to provide a digital audio encoding apparatus and a digital audio encoding apparatus which can provide an audio quality or a compression rate superior to that of the conventional digital audio encoding system and can be applied to a conventional digital audio encoding system for performing compression in a spectral region To provide.

1 is a block diagram illustrating a digital audio encoding apparatus according to a first embodiment of the present invention.

FIG. 2 is a detailed block diagram of the spectrum modeling unit in FIG. 1,

FIG. 3 is a detailed block diagram of the Bark-band gain calculator in FIG. 2,

4 is a block diagram illustrating a digital audio decoding apparatus according to a second embodiment of the present invention.

FIG. 5 is a detailed block diagram of the output audio synthesis unit in FIG. 4,

FIG. 6 is a detailed block diagram of the spectrum synthesis unit in FIG. 5,

FIG. 7 is a conceptual diagram to which a method of calculating a gain of a gain calculator is applied in FIG. 3,

FIG. 8 is a conceptual diagram in which a method of restoring the spectrum of the spectrum combining unit is applied in FIG. 5,

9 is a block diagram illustrating a digital audio encoding / decoding apparatus according to a third embodiment of the present invention.

Description of the Related Art [0002]

100: down-sampling unit 200: audio coding unit

300: Spectrum modeling unit 400: Multiplexing unit

500: demultiplexing unit 600: audio decoding unit

700: Gain extraction unit 800: Output audio synthesis unit

310: frequency conversion unit 320: upsampling unit

330: Bark-band gain calculation unit 340: Quantization unit

331: First Bark Band Transformation Unit 332: Second Bark Band Transformation Unit

333: psychoacoustic modeling unit 334: first spectral band selection unit

335: second spectrum band selection unit 336: gain calculation unit

810: Upsampling unit 820: Spectrum synthesizing unit

830: Time base conversion unit 821: Spectrum selection unit

822: high frequency band generating section 823: output spectrum synthesizing section

10: encoding unit 20: decoding unit

11: down-sampling unit 12: audio coding unit

13: Spectrum modeling unit 14: Multiplexing unit

21: demultiplexing unit 22: audio decoding unit

23: Gain extraction unit 24: Output audio synthesis unit

According to an aspect of the present invention, there is provided a method of decoding a digital audio signal, the method comprising: sampling a sampling frequency of an input audio signal using a frequency higher than a sampling frequency to be encoded, down- And the spectrum value for the frequency region which can not be encoded is the spectrum of the input audio signal before down sampling in the Bark-Band band and the energy ratio of the spectrum of the compressed result to The spectrum of the low frequency region is copied and restored in the decoding process.

Here, the Bark band is a linear band which is converted into a linear band recognized by humans. The Bark band increases linearly at a low frequency and increases logarithmically toward a high frequency. In the present invention, Since the band is a high frequency band, the number of high frequency bands can be reduced by using the bark band.

In addition, multiplying the low-frequency band by the energy ratio and then copying and restoring the audio spectrum itself is characterized by harmonics resonating with the fundamental frequency, so that the high-frequency band is not completely recovered, but the spectrum region is expanded A better sound quality can be obtained, and the value of the energy ratio for spectrum recovery in the encoded bit stream is added as an additional bit stream.

According to an aspect of the present invention, there is provided a digital audio encoding apparatus including:

A down-sampling unit for receiving a digital audio signal and down-sampling the digital audio signal by a factor of 1 / N;

An audio encoding unit for encoding the downsampled audio signal output from the downsampling unit using a digital audio encoding scheme;

A spectral modeling unit that receives a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal output from the audio encoding unit and uses the gain value in a bark band between the two spectrums to model a frequency band that can not be encoded, part; And

A spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit and an encoded bitstream output from the audio encoder, and multiplexing the encoded bitstream to maintain compatibility with the digital audio encoding standard, thereby outputting a final bitstream; .

The spectral modeling unit,

A frequency converter for receiving a digital audio signal and converting the digital audio signal into a frequency band to generate a original sound spectrum;

An up-sampling unit for up-sampling an encoding spectrum output from the audio encoding unit N times to generate a spectrum having a same frequency band as the original sound spectrum;

A Bark-band gain calculator for calculating, as a gain value, an energy ratio of a low-frequency band of a coding spectrum up-sampled in a Bark band and a high-frequency band of the original sound spectrum in order to model a spectrum value of a frequency band that can not be expressed in a coding spectrum; And

And a quantizer for receiving a gain value output from the Bark Band gain calculator and quantizing the gain value to generate a spectrum compensator bitstream and outputting the bitstream to the multiplexer.

The Bark-band gain calculator includes:

A first Bark band transformer for converting the original sound spectrum band output from the frequency converter into a Bark band;

A psychoacoustic modeling unit for removing a spectral component that does not affect the sound quality by using a psychoacoustic model as the original sound spectrum converted into a bark band output from the first Bark Band transform unit;

A first spectral band selection unit for selecting a spectral band of a part that can not be encoded in the encoding process in the spectrum output from the psychoacoustic modeling unit;

A second Bark Band transformer for transforming a band of the upsampled encoded spectrum into a Bark band;

A second spectrum band selector for selecting a low frequency band in an upsampled encoding spectrum converted into a bark band output from the second bark band converter; And

And a gain calculator for calculating an energy in a bark band between two spectral bands output from the first and second spectral band selection units and calculating a gain value and outputting the gain value to the quantization unit.

According to another aspect of the present invention, there is provided a digital audio decoding apparatus comprising:

A demultiplexing unit for receiving the bitstream and separating the encoded bitstream and the spectrum-compensated bitstream separately encoded by a conventional digital audio encoding scheme;

An audio decoding unit for receiving and decoding a coded bit stream output from the demultiplexing unit to generate a decoded spectrum;

A gain extraction unit for receiving and decoding a spectrum compensator bitstream output from the demultiplexer to calculate and output a gain value in a bark band; And

An output audio synthesis unit that receives the decoded spectrum output from the audio decoding unit, restores the spectrum by expanding the frequency spectrum before down sampling using the gain value output from the gain extraction unit, converts the spectrum into a time base signal, .

Wherein the output audio synthesis unit comprises:

An up-sampling unit for receiving the decoded spectrum output from the audio decoding unit and up-sampling the decoded spectrum;

A spectral synthesizer for receiving a upsampled decoded spectrum of a low frequency band output from the upsampling unit and multiplying the upsampled decoded spectrum by a gain output from the gain extraction unit to generate a transform spectrum in which a high frequency band is reconstructed and outputting the converted spectrum; And

And a time axis converting unit for receiving the transformed reconstructed spectrum output from the spectrum synthesizing unit and inversely transforming the transformed spectrum into a time axis to generate an output audio signal.

The spectral synthesizer may comprise:

A spectral selector for receiving a upsampled decoded spectrum output from the upsampling unit and selecting and outputting a spectrum of a low frequency band;

A high frequency band generating unit which receives a spectrum of a selected low frequency band outputted from the spectrum selecting unit and multiplies the spectrum of the selected low frequency band by a gain value outputted from the gain extracting unit and replaces it with a spectrum of a high frequency band; And

And a spectral combining unit for receiving the spectrum of the recovered high frequency band outputted from the high frequency band generating unit and merging the upsampled decoded spectrum output from the up sampling unit to generate an output spectrum and outputting the output spectrum to the time axis converting unit.

According to another aspect of the present invention, there is provided a digital audio encoding / decoding apparatus including:

A down-sampling unit for receiving a digital audio signal and down-sampling the digital audio signal by a factor of 1 / N;

An audio encoding unit for encoding the downsampled audio signal output from the downsampling unit using a digital audio encoding scheme;

A spectral modeling unit that receives a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal output from the audio encoding unit and uses the gain value in a bark band between the two spectrums to model a frequency band that can not be encoded, part;

A spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit and an encoded bitstream output from the audio encoder, and multiplexing the encoded bitstream to maintain compatibility with the digital audio encoding standard, thereby outputting a final bitstream; part;

A demultiplexing unit for receiving the bitstream and separating the encoded bitstream and the spectrum-compensated bitstream separately encoded by a conventional digital audio encoding scheme;

An audio decoding unit for receiving and decoding a coded bit stream output from the demultiplexing unit to generate a decoded spectrum;

A gain extraction unit for receiving and decoding a spectrum compensator bitstream output from the demultiplexer to calculate and output a gain value in a bark band; And

An output audio synthesis unit that receives the decoded spectrum output from the audio decoding unit, restores the spectrum by expanding the frequency spectrum before down sampling using the gain value output from the gain extraction unit, converts the spectrum into a time base signal, .

According to another aspect of the present invention, there is provided a digital audio encoding method,

A digital audio encoding method for receiving a digital audio signal, compressing the digital audio signal, and outputting the digital audio signal as a bitstream,

A first step of receiving an input digital audio signal and down-sampling the input digital audio signal by a factor of 1 / N;

A second step of generating a bitstream by encoding the downsampled audio signal using a digital audio encoding scheme;

A third step of receiving a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal and modeling a frequency band that can not be encoded using a gain value in a bark band between two spectrums; And

And a fourth step of receiving the spectrum-compensated bitstream and the audio-coded bitstream obtained by encoding the gain value and multiplexing the received bitstream so as to maintain compatibility with the digital audio encoding standard.

According to another aspect of the present invention, there is provided a digital audio decoding method comprising:

A digital audio decoding method for receiving a compressed bitstream and restoring the original digital audio signal,

A first step of separating an input bit stream into an encoded bit stream and a spectrum compensated bit stream which are encoded by a conventional digital audio encoding method;

A second step of receiving and decoding the encoded bit stream to generate a decoded spectrum;

A third step of receiving and decoding the spectrum compensator bitstream and calculating a gain value in a bark band;

A fourth step of receiving the decoded spectrum and expanding the decoded spectrum to a frequency band before downsampling using the gain value to recover a spectrum; And

And a fifth step of converting the recovered spectrum into a time-axis signal.

Hereinafter, a first preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

As shown in FIG. 1, the digital audio encoding apparatus according to the first embodiment of the present invention includes:

An apparatus for encoding an input digital audio signal includes a downsampling unit 100, an audio encoding unit 200, a spectrum modeling unit 300, and a multiplexing unit 400.

The downsampling unit 100 receives the digital audio signal and downsamples the digital audio signal by a factor of 1 / N and outputs the downsampled signal.

The audio encoding unit 200 encodes the downsampled audio signal output from the downsampling unit 100 using a digital audio encoding scheme.

The spectral modeling unit 300 receives a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal output from the audio encoding unit 200 and uses the gain value in a bark band between the two spectrums, And models the frequency band that can not be encoded.

The multiplexer 400 receives the spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit 300 and the encoded bitstream output from the audio encoder 200, So as to output the final bitstream.

2, the spectrum modeling unit 300 includes a frequency transform unit 310, an upsampling unit 320, a bark band gain calculation unit 330, and a quantization unit 340.

The frequency converting unit 310 receives the digital audio signal, converts the digital audio signal into a frequency band, and generates an original sound spectrum and outputs the original sound spectrum to the upsampling unit 320.

The up-sampling unit 320 up-samples the encoded spectrum output from the audio encoding unit 200 by N times and outputs a spectrum having the same frequency band as the original sound spectrum.

The Bark-band gain calculator 330 calculates the energy ratio of the low-frequency band of the encoded spectrum up-sampled in the Bark band and the high-frequency band of the original sound spectrum as a gain value in order to model the spectrum value of the frequency band that can not be expressed in the encoding spectrum And outputs it to the quantization unit 340.

The quantizer 340 receives the gain value output from the Bark-band gain calculator 330 and quantizes the gain value to generate a spectrum-compensated bitstream, and outputs the bitstream to the multiplexer 400.

3, the Bark Band gain calculator 330 includes a first Bark Band transformer 331, a second Bark Band transformer 3320, a psychoacoustic modeler 333, 1 spectral band selection unit 335, a second spectrum band selection unit 334, and a gain calculation unit 336.

The first Bark band transformer 331 transforms the original sound spectrum band output from the frequency transformer 310 into a Bark band and the second Bark band transformer 332 transforms the band of the up- Bark band.

The psychoacoustic modeling unit 333 removes a spectral component that does not affect the sound quality using the psychoacoustic model and outputs the original sound spectrum converted into the bark band output from the first Bark Band transformer 331 .

The first spectral band selector 335 selects a spectral band of a part of the spectrum output from the psychoacoustic modeling unit 333 that can not be encoded in the encoding process.

The second spectrum band selector 334 selects a low frequency band from the upsampled encoded spectrum converted from the bark band output from the second bark band transformer 332 and outputs the low frequency band.

The gain calculator 336 calculates the gain in the Bark band between the two spectral bands output from the first spectral band selector 333 and the second spectral band selector 334, (340).

The operation of the first embodiment of the present invention is as follows.

1, the downsampling unit 100 is connected to the audio encoding unit 200 to downsample an input digital audio signal sampled at a sampling frequency Fs by 1 / N.

The audio encoding unit 200 receives the downsampled audio signal output from the downsampling unit 100 and encodes the downsampled audio signal using a conventional digital audio encoding method such as AAC or MP3.

Then, the audio coding unit 200 outputs the coding spectrum to the spectrum modeling unit 300, and outputs the coding bitstream to the multiplexing unit 400. [

The spectrum modeling unit 300 receives the encoded spectrum output from the audio encoding unit 200 and the input digital audio signal to model a spectrum of a high frequency band that can not be expressed by the audio encoding unit 100, And generates a spectrum-compensated bitstream using the generated bitstream and outputs the generated bitstream to the multiplexer 400.

The multiplexer 400 receives the encoded bit stream output from the audio encoder 100 and the spectral compensator bit stream output from the spectral modeler 300 and decodes the encoded bit stream, And outputs the generated output bit stream so as to maintain compatibility.

This is because the spectral compensator bitstream is inserted as additional information at the end of each frame of the encoded bitstream so that the decoder of the existing digital audio encoding system does not recognize and process the additional data stream, do.

Hereinafter, the operation of the spectrum modeling unit 300 will be described in detail with reference to FIG. 2

The frequency converter 310 of the spectrum modeling unit 300 receives a digital audio signal as a time axis signal, converts the digital audio signal into a frequency signal using FFT or the like, and outputs the frequency signal.

The upsampling unit 320 upsamples the encoding spectrum output from the audio encoding unit 100 shown in FIG. 1 to N times, and generates an upsampled encoding spectrum.

The Bark-band gain calculator 330 receives the original sound spectrum output from the frequency converter 310 and the encoded spectrum output from the up-sampling unit 320, and outputs a low frequency band of a coding spectrum up- The energy ratio of the high frequency band of the original sound spectrum is calculated as a gain value.

The quantizer 340 quantizes the gain value output from the Bark-band gain calculator 330 to generate a spectrum-compensated bitstream and outputs the bitstream to the multiplexer 400.

3 is a detailed block diagram of the BarkGain gain calculator 330, and the operation of the BarkBand gain calculator 330 will be described with reference to FIG.

The first Bark band transformer 331 receives the original sound spectrum output from the frequency transform unit 310 of FIG. 2, converts the original sound spectrum into a Bark band, and generates and outputs a first transform spectrum.

The psychoacoustic modeling unit 333 receives the first transform spectrum output from the first Bark band transform unit 331 and applies a psychoacoustic model modeled by a human perception characteristic of the sound to affect the sound quality And outputs the frequency component.

The first spectrum band selector 335 selects high frequency bands that can not be expressed by the audio coding unit 100 of FIG. 1 in the first conversion spectrum outputted from the psychoacoustic modeling unit 302, The energy is calculated and output.

On the other hand, the second Bark band transformer 332 receives the upsampled encoded spectrum output from the upsampling unit 320 of FIG. 2, converts it into a Bark band, and outputs the Bark band.

The second spectrum band selector 334 selects low frequency bands using basic information for recovering the high frequency band in the second converted spectrum converted from the bark band outputted from the second bark band transformer 332 And calculates and outputs an energy value for the band.

The gain calculator 336 calculates the ratio of the energy values of the bands output from the first spectral band selector 335 and the spectral energy value output from the second spectral band selector 334, And outputs it to the quantization unit 340.

As described above, in the first embodiment of the present invention, the encoding compression ratio can be improved by compensating for a band that can not be encoded through a gain value using a bark band.

4 and 6 show a digital audio decoding apparatus for decoding an audio signal by decoding the coded signal.

Hereinafter, a second preferred embodiment of the present invention will be described with reference to the accompanying drawings.

As shown in FIG. 4, the digital audio decoding apparatus according to the embodiment of the present invention includes:

An audio multiplexing unit 500, an audio decoding unit 600, a gain extracting unit 700, and an output audio synthesizing unit 800.

The demultiplexer 500 demultiplexes the coded bit stream and the spectrum compensator bit stream, which are coded by the conventional digital audio encoding scheme, into a bit stream and outputs the bit stream.

The audio decoding unit 600 receives and decodes a coded bit stream output from the demultiplexing unit 500 to generate a decoded spectrum. The gain extracting unit 700 extracts the coded bit stream outputted from the demultiplexing unit 500 And outputs the calculated gain value in the bark band.

The output audio synthesis unit 800 receives the decoded spectrum output from the audio decoding unit 600 and uses the gain value output from the gain extraction unit 700 to expand it to a frequency band before downsampling Restores the spectrum, and converts the spectrum into a time-base signal and outputs it.

5, the output audio synthesis unit 800 includes an upsampling unit 810, a spectrum synthesis unit 820, and a time axis conversion unit 830.

The upsampling unit 810 receives the decoded spectrum output from the audio decoding unit 600 and upsamples the decoded spectrum N times and outputs the decoded spectrum. The spectrum synthesizing unit 820 generates a low frequency Sampled decoded spectrum of the frequency band of the received signal and multiplies the gain value output from the gain extraction unit 700 to generate a transform spectrum in which the high frequency band is reconstructed and outputs the converted spectrum.

The time-base converter 830 receives the transformed spectrum output from the spectrum synthesizer 820 and inversely transforms the spectrum into a time axis to generate and output an output audio signal.

6, the spectrum synthesis unit 820 includes a spectrum selection unit 821, a high-frequency band generation unit 822, and an output spectrum synthesis unit 823.

The spectrum selector 821 receives the upsampled decoded spectrum output from the upsampling unit 810 and selects and outputs a spectrum of a low frequency band.

The high frequency band generating unit 822 receives the spectrum of the selected low frequency band outputted from the spectrum selecting unit 821 and multiplies the spectrum of the selected low frequency band by the gain value output from the gain extracting unit 700 and replaces it with the spectrum of the high frequency band.

The output spectral synthesizer 820 receives the spectrum of the reconstructed high frequency band output from the high frequency band generator 822 and merges the upsampled decoded spectrums output from the up-sampling unit 810 to obtain an output spectrum And outputs it to the time base conversion unit 830.

The operation of the second embodiment of the present invention as described above is as follows.

4, the demultiplexing unit 500 of the digital audio decoding apparatus receives a bitstream, separates the encoded bitstream generated by the conventional digital audio encoding method such as AAC and MP3, and the spectrum-compensated bitstream And outputs it.

Then, the audio decoding unit 600 decodes the separated encoded bit stream outputted from the demultiplexing unit 500 to generate a decoded spectrum, and outputs the decoded spectrum to the output audio synthesizing unit 800.

The gain extracting unit 700 analyzes the separated spectrum compensator bit stream output from the demultiplexer 500 and generates a gain value necessary for synthesizing the output audio signal.

The output audio synthesis unit 800 reconstructs the decoded spectrum decoded by the audio decoding unit 600 using the gain values extracted from the gain extraction unit 700 and reconstructs the spectrum of the high frequency band using the recovered spectrum as a time axis And generates and outputs a final output audio signal.

Hereinafter, the output audio synthesizer 800 will be described in detail with reference to FIG.

The upsampling unit 810 receives the decoded spectrum generated by the audio decoding unit 600 of FIG. 1, and upsamples the decoded spectrum by N times to generate an upsampled decoded spectrum.

The spectral combining unit 820 selects a low frequency band to be extended in the high frequency band from the upsampled decoding spectrum output from the upsampling unit 810 and then outputs the gain value calculated by the gain extracting unit 700 of FIG. And the spectrum of the high frequency band is restored to generate the converted spectrum.

The time base conversion unit 830 converts the converted spectrum generated by the spectrum synthesis unit 820 into a time base, and generates and outputs an output audio signal.

A detailed block diagram of the spectrum synthesizer 820 is shown in FIG.

In FIG. 6, the spectrum selector 821 selects a low frequency band used as a base band for recovering a high frequency band signal in the upsampled decoding spectrum generated by the upsampling unit 810 of FIG.

The high frequency band generating unit 822 multiplies the low frequency band spectrum output from the spectrum selecting unit 821 by the gain values extracted by the gain extracting unit 700 shown in FIG. 4 to restore a spectrum of a high frequency band to be restored do.

The output spectrum synthesis unit 823 merges the upsampled decoded spectrum generated by the upsampling unit 810 of FIG. 5 and the high-frequency band spectrum reconstructed by the high-frequency band generation unit 822 to generate an output spectrum, And outputs it to the time-base conversion unit 830.

By doing so, the digital audio decoding apparatus according to the embodiment of the present invention can restore the encoded audio signal to the original signal band.

7 and 8 illustrate a process of downsampling input audio with an N value of 2, that is, a half time, and a process of modeling and restoring the spectrum of the input bit stream, have.

FIG. 7 is a conceptual diagram showing a method of calculating the gain of the gain calculator 336 of FIG. 3. Since the spectrum of the encoded audio signal is down-sampled and coded by a factor of 1/2, it is less than half of the spectrum of the input audio signal Is represented by only the component of < RTI ID = 0.0 >

Therefore, the components for the first group (Group A) and the second group (Group B) can not be expressed. In the present invention, the low frequency region of the spectrum of the encoded audio signal represented by the third group (Group C) The energy ratio between the group A and the group B is calculated and the first gain value Gain 1 and the second gain value Gain 2 are extracted as the gain value.

The calculated first gain value (Gain 1) and second gain value (Gain 2) are quantized and multiplexed with the encoded bit string.

FIG. 8 shows a process of restoring a high frequency component using the gain calculated in FIG.

A frequency band is expanded by selecting a low frequency band from the spectrum of the restored audio signal, multiplying the gain value, and then merging the frequency band into a high frequency band.

Since most of the information in the audio signal is located in the low frequency band near 1 KHz and the high frequency band is known to affect the enhancement of the clarity of the sound quality except the special signal, if the high frequency is restored by the above method, There is an effect of increasing the clarity of the audio signal while reducing the number of bits for representation.

In FIGS. 7 and 8, Fs represents the sampling frequency of the downsampled signal, and the interval indicated by the dotted line represents the Bark band.

On the other hand, FIG. 9 shows a digital audio encoding / decoding device in which the digital audio encoding apparatus described in the first embodiment and the digital audio decoding apparatus described in the second embodiment are integrated.

Hereinafter, a third embodiment of the present invention will be described with reference to FIG.

9, in the digital audio encoding / decoding apparatus according to the third embodiment of the present invention,

An encoder 10, and a decoder 20.

The encoder 10 includes a downsampling unit 11, an audio encoding unit 12, a spectrum modeling unit 13, and a multiplexing unit 14.

The encoder 20 includes a demultiplexer 21, an audio decoder 22, a gain extractor 23, and an output audio synthesizer 24.

The downsampling unit 11 of the encoder 10 receives the digital audio signal and downsamples it by a factor of 1 / N and outputs the downsampled signal.

The audio encoding unit 12 encodes the downsampled audio signal output from the downsampling unit using a digital audio encoding scheme and outputs the encoded downsampled audio signal.

The spectral modeling unit 13 receives the spectrum of the input digital audio signal and the downsampled signal output from the audio encoding unit 12 and uses the gain value in the bark band between the spectrums of the compressed signal, And models the frequency band that can not be encoded.

The multiplexer 14 receives the spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit 13 and the encoded bitstream output from the audio encoder 12, So as to output the final bitstream.

The demultiplexing unit 21 of the decoding unit 20 receives the bitstream and separates the encoded bitstream and the spectrum-compensated bitstream into a bitstream and a spectrum-compensated bitstream.

The audio decoding unit 22 receives and decodes an encoded bit stream output from the demultiplexing unit 21 to generate a decoded spectrum.

The gain extracting unit 23 receives and decodes a spectral compensating unit bit stream output from the demultiplexing unit 21, and calculates and outputs a gain value in a bark band.

The output audio synthesis unit 24 receives the decoded spectrum output from the audio decoding unit 22 and uses the gain value output from the gain extraction unit 23 to expand the spectrum to a frequency band before downsampling, And converts the spectrum into a time-base signal and outputs it.

The operation of the third embodiment as described above is as follows.

The downsampling unit 11 of the encoder 10 is connected to the audio encoding unit 12 and downsamples the input digital audio signal sampled at the sampling frequency Fs by 1 / N.

The audio encoding unit 12 receives the downsampled audio signal output from the downsampling unit 11 and encodes the downsampled audio signal using a conventional digital audio encoding scheme such as AAC or MP3.

Then, the audio coding unit 12 outputs the coding spectrum to the spectrum modeling unit 13, and outputs the coding bit stream to the multiplexing unit 14. [

The spectrum modeling unit 13 receives a coding spectrum output from the audio coding unit 12 and the input digital audio signal and models a spectrum of a high frequency band that can not be expressed by the audio coding unit 11, And generates a spectrum-compensated bitstream using the generated bitstream and outputs the bitstream to the multiplexer 14. [

The multiplexer 14 receives the encoded bit stream output from the audio encoder 11 and the spectrum compensator bit stream output from the spectrum modeler 13 and decodes And outputs the generated output bit stream so as to maintain compatibility.

This is because the spectral compensator bitstream is inserted as additional information at the end of each frame of the encoded bitstream so that the decoder of the existing digital audio encoding system does not recognize and process the additional data stream, do.

The demultiplexer 21 of the decoder 20 receives the bitstream, separates the encoded bitstream and the spectrum-compensated bitstream generated by the conventional digital audio encoding method such as AAC and MP3, do.

Then, the audio decoding unit 22 decodes the separated encoded bit stream outputted from the demultiplexing unit 21 to generate a decoded spectrum, and outputs the decoded spectrum to the output audio synthesizing unit 800.

The gain extracting unit 23 analyzes the separated spectrum compensator bit stream output from the demultiplexer 21 and generates a gain value necessary for synthesizing the output audio signal.

The output audio synthesis unit 24 reconstructs the decoded spectrum decoded by the audio decoding unit 22 using the gain value extracted by the gain extraction unit 23 and reconstructs the spectrum of the high frequency band using the reconstructed spectrum as a time axis And generates and outputs a final output audio signal.

The specific configurations and operations of the encoder 10 and the decoder 20 are the same as those described in the first and second embodiments. However, since the configurations and functions of the encoder 10 and the decoder 20 are different from each other, Explanations are omitted in order to avoid redundancy.

As described in the first, second, and third embodiments, the present invention can be applied to a method of encoding an input digital audio signal by supplementing a non-coding band with a gain value using a bark band And by restoring using the gain value when restoring, the compression ratio can be improved as compared with the conventional method.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims. It can be understood that it is possible.

Therefore, the present invention operating as described above can exhibit the same effect as compression using a higher sampling frequency while lowering the sampling frequency of the input audio signal for coding at a low bit rate of 32 Kbps or less.

The present invention which operates as described above has the effect of greatly improving the compression rate while maintaining the sound quality of a conventional digital audio compression method such as MP3 or AAC.

In addition, since the present invention operates in the above-described manner using the encoding spectrum generated during the conventional digital audio encoding process, the additional computation amount for implementing the present invention is small, and at a low bit rate of 32 Kbps or less, It is possible to obtain better sound quality at the same compression rate.

Therefore, the present invention, which operates as described above, can compress a small amount of data while maintaining a similar sound quality, thereby increasing the efficiency of the storage medium and increasing the transmission efficiency.

In addition, since the spectrum compensator bitstream is transmitted / stored in a bit stream obtained by encoding the downsampled audio signal, compatibility with the conventional digital audio compression method can be maintained.

Claims (9)

A down-sampling unit for receiving a digital audio signal and down-sampling the digital audio signal by a factor of 1 / N; An audio encoding unit for encoding the downsampled audio signal output from the downsampling unit using a digital audio encoding scheme; A spectral modeling unit that receives a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal output from the audio encoding unit and uses the gain value in a bark band between the two spectrums to model a frequency band that can not be encoded, part; And A spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit and an encoded bitstream output from the audio encoder, and multiplexing the encoded bitstream to maintain compatibility with the digital audio encoding standard, thereby outputting a final bitstream; And a digital audio encoding unit. The apparatus of claim 1, wherein the spectrum modeling unit comprises: A frequency converter for receiving a digital audio signal and converting the digital audio signal into a frequency band to generate a original sound spectrum; An up-sampling unit for up-sampling the encoded spectrum output from the audio encoding unit by N times to generate a spectrum having the same frequency band as the original sound spectrum; A Bark-band gain calculator for calculating, as a gain value, an energy ratio of a low-frequency band of a coding spectrum up-sampled in a Bark band and a high-frequency band of the original sound spectrum in order to model a spectrum value of a frequency band that can not be expressed in a coding spectrum; And And a quantizer configured to receive and quantize a gain value output from the Bark-band gain calculator to generate a spectrum-compensated bitstream and to output the bit-stream to the multiplexer. 3. The apparatus of claim 2, wherein the Bark- A first Bark band transformer for converting the original sound spectrum band output from the frequency converter into a Bark band; A psychoacoustic modeling unit that removes spectral components that do not affect sound quality using a psychoacoustic model, the original sound spectrum converted into a bark band output from the first Bark Band transform unit; A first spectral band selection unit for selecting a spectral band of a part that can not be encoded in the encoding process in the spectrum output from the psychoacoustic modeling unit; A second Bark Band transformer for transforming a band of the upsampled encoded spectrum into a Bark band; A second spectrum band selector for selecting a low frequency band in an upsampled encoding spectrum converted into a bark band output from the second bark band converter; And And a gain calculator for obtaining energy in a bark band between two spectral bands output from the first and second spectral band selection units, and calculating a gain value and outputting the gain values to the quantization unit. A demultiplexing unit for receiving the bitstream and separating the encoded bitstream and the spectrum-compensated bitstream separately encoded by a conventional digital audio encoding scheme; An audio decoding unit for receiving and decoding a coded bit stream output from the demultiplexing unit to generate a decoded spectrum; A gain extraction unit for receiving and decoding a spectrum compensator bitstream output from the demultiplexer to calculate and output a gain value in a bark band; And An output audio synthesis unit that receives the decoded spectrum output from the audio decoding unit, restores the spectrum by expanding the frequency spectrum before down sampling using the gain value output from the gain extraction unit, converts the spectrum into a time base signal, And a digital audio decoding unit. 5. The audio decoding apparatus according to claim 4, An up-sampling unit for receiving the decoded spectrum output from the audio decoding unit and up-sampling the decoded spectrum; A spectral synthesizer for receiving a upsampled decoded spectrum of a low frequency band output from the upsampling unit and multiplying the upsampled decoded spectrum by a gain output from the gain extraction unit to generate a transform spectrum in which a high frequency band is reconstructed and outputting the converted spectrum; And And a time axis converting unit for receiving the converted reconstructed spectrum output from the spectrum synthesizing unit and generating an output audio signal by inversely converting the spectrum into a time axis. 6. The apparatus of claim 5, A spectral selector for receiving a upsampled decoded spectrum output from the upsampling unit and selecting and outputting a spectrum of a low frequency band; A high frequency band generating unit which receives a spectrum of a selected low frequency band outputted from the spectrum selecting unit and multiplies the spectrum of the selected low frequency band by a gain value outputted from the gain extracting unit and replaces it with a spectrum of a high frequency band; And And a spectrum synthesizer for generating an output spectrum by combining the upsampled decoded spectrum outputted from the upsampling section and receiving the spectrum of the reconstructed high frequency band outputted from the high frequency band generation section and outputting the output spectrum to the time base conversion section, Decoding device. A down-sampling unit for receiving a digital audio signal and down-sampling the digital audio signal by a factor of 1 / N; An audio encoding unit for encoding the downsampled audio signal output from the downsampling unit using a digital audio encoding scheme; A spectral modeling unit that receives a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal output from the audio encoding unit and uses the gain value in a bark band between the two spectrums to model a frequency band that can not be encoded, part; A spectral compensator bitstream obtained by encoding the gain value output from the spectral modeling unit and an encoded bitstream output from the audio encoder, and multiplexing the encoded bitstream to maintain compatibility with the digital audio encoding standard, thereby outputting a final bitstream; part; A demultiplexing unit for receiving the bitstream and separating the encoded bitstream and the spectrum-compensated bitstream separately encoded by a conventional digital audio encoding scheme; An audio decoding unit for receiving and decoding a coded bit stream output from the demultiplexing unit to generate a decoded spectrum; A gain extraction unit for receiving and decoding a spectrum compensator bitstream output from the demultiplexer to calculate and output a gain value in a bark band; And An output audio synthesis unit that receives the decoded spectrum output from the audio decoding unit, restores the spectrum by expanding the frequency spectrum before down sampling using the gain value output from the gain extraction unit, converts the spectrum into a time base signal, And a digital audio encoding / decoding device. A digital audio encoding method for receiving a digital audio signal, compressing the digital audio signal, and outputting the digital audio signal as a bitstream, A first step of receiving an input digital audio signal and down-sampling the input digital audio signal by a factor of 1 / N; A second step of generating a bitstream by encoding the downsampled audio signal using a digital audio encoding scheme; A third step of receiving a spectrum of an input digital audio signal and a spectrum of a downsampled compressed signal and modeling a frequency band that can not be encoded using a gain value in a bark band between two spectrums; And And a fourth step of receiving the spectrum-compensated bitstream and the audio-encoded bitstream obtained by coding the gain values and multiplexing the received bitstreams so as to maintain compatibility with the digital audio encoding standard. A digital audio decoding method for receiving a compressed bitstream and restoring the original digital audio signal, A first step of separating an input bit stream into an encoded bit stream and a spectrum compensated bit stream which are encoded by a conventional digital audio encoding method; A second step of receiving and decoding the encoded bit stream to generate a decoded spectrum; A third step of receiving and decoding the spectrum compensator bitstream and calculating a gain value in a bark band; A fourth step of receiving the decoded spectrum and expanding the decoded spectrum to a frequency band before downsampling using the gain value to recover a spectrum; And And a fifth step of converting the recovered spectrum into a time-base signal.