KR101393300B1 - Method and Apparatus for decoding audio/speech signal - Google Patents

Abstract

The present invention relates to a method of coding an audio / speech signal, wherein a length of a frame, which is a processing unit of an input signal, is variably determined according to an attack position of an input signal, and an input signal is transformed into a frequency domain If it is determined that the divided subband signal is encoded in the frequency domain, the subband signal is encoded in the frequency domain. If it is determined that the divided subband signal is encoded in the time domain, The signal is encoded in the time domain by inverse transforming the signal into the time domain, thereby effectively encoding the audio / speech signal by adjusting the time resolution and the frequency resolution.

Description

[0001] The present invention relates to a method and apparatus for decoding audio / speech signals,

1 is a block diagram illustrating an audio / speech signal encoding apparatus according to an embodiment of the present invention.

FIG. 2 is a graph illustrating a frame to be adjusted in an audio / speech signal encoding apparatus according to an embodiment of the present invention. Referring to FIG.

FIG. 3 is a graph illustrating encoding domains for each frame and band in an audio / speech signal encoding apparatus according to an exemplary embodiment of the present invention. Referring to FIG.

4 is a block diagram illustrating an audio / speech signal decoding apparatus according to an embodiment of the present invention.

5 is a flowchart illustrating a method of encoding an audio / speech signal according to an embodiment of the present invention.

6 is a flowchart illustrating a method of decoding an audio / speech signal according to an embodiment of the present invention.

The present invention relates to a method and apparatus for encoding and decoding audio and speech signals.

Conventional codecs are classified into a speech codec and an audio codec. The speech codec encodes or decodes a signal corresponding to a frequency band ranging mainly from 50 Hz to 7 kHz using a speech utterance model. Such a speech codec generally encodes and decodes parameters representative of a speech signal by modeling vocal cords and syllables. The audio codec applies a psychoacoustic model like HE-AAC to encode or decode signals corresponding to a frequency range from 0 Hz to 24 kHz. The audio codec performs coding and decoding by omitting low-sensitivity signals using human auditory characteristics.

However, the speech codec is suitable for encoding or decoding a speech signal, but sound quality may be degraded in encoding or decoding an audio signal. The audio codec is excellent in compression effect when encoding or decoding an audio signal, but efficiency of compressing a signal in encoding or decoding a voice signal may be lowered. Therefore, there is a need for a method and apparatus capable of improving compression efficiency and sound quality when encoding or decoding a speech signal, an audio signal, or a mixed signal of speech and audio, respectively.

SUMMARY OF THE INVENTION The present invention is directed to a method and apparatus for encoding an audio / speech signal that can improve compression efficiency and sound quality by reflecting characteristics of an input signal.

According to another aspect of the present invention, there is provided a method and apparatus for decoding an audio / speech signal capable of improving compression efficiency and sound quality by reflecting characteristics of an input signal.

According to another aspect of the present invention, there is provided a method of encoding an audio / speech signal, the method comprising: variably determining a length of a frame, which is a processing unit of an input signal, according to an attack position of the input signal; Converting the input signal into a frequency domain and dividing the input signal into subbands; and if the divided subband signal is determined to be encoded in the frequency domain, encoding the subband signal in a frequency domain, If the signal of the subband is determined to be encoded in the time domain, inverse transforming the signal of the subband into the time domain and encoding in the time domain.

According to another aspect of the present invention, there is provided a method for processing an input signal, the method comprising: varying a length of a frame, which is a processing unit of an input signal, according to an attack position of the input signal; Dividing a signal of the subband into a frequency domain and a frequency domain of the signal of the subband when the signal of the subband is determined to be encoded in the frequency domain; And encoding the audio / speech signal in a time domain by inversely transforming the signal of the subband into a time domain.

According to another aspect of the present invention, there is provided an apparatus for encoding an audio / speech signal, the apparatus including: a frame determination unit configured to variably determine a length of a frame, which is a processing unit of an input signal, according to an attack position of the input signal; A domain converter for converting the input signal into a frequency domain and dividing the input signal into subbands for each of the subbands, and determining whether to encode the signals of the subbands in the frequency domain or the time domain for each of the divided subbands, A domain determination unit, a domain inversion unit that inversely transforms a signal of a subband determined to be encoded in the time domain into a time domain, and a subband determination unit that encodes the signal of the inversely transformed subband in a time domain and determines a subband The frequency domain It includes stand encoding unit for encoding.

According to another aspect of the present invention, there is provided a method of decoding an audio / speech signal, the method comprising: determining a domain coded by a frame and a subband of a coded signal; Decoding the signal in the frequency domain, decoding the signal in the frequency domain, and synthesizing the decoded signal in the time domain and the decoded signal in the frequency domain, do.

According to another aspect of the present invention, there is provided a method of decoding a signal, comprising the steps of: determining a domain encoded by a frame and a subband of the encoded signal; decoding the signal determined to be encoded in the time domain in a time domain; Decoding a signal in a frequency domain, and synthesizing a signal decoded in the time domain and a signal decoded in the frequency domain and inverse-transforming the decoded signal into a time domain, the method comprising: And a computer-readable recording medium recorded thereon.

According to another aspect of the present invention, there is provided an apparatus for decoding an audio / speech signal, the apparatus including: a determination unit for determining a domain encoded by a frame and a subband of a coded signal; A decoding unit for decoding the signal decoded in the time domain and decoded in the frequency domain, and a decoding unit for decoding the signal decoded in the time domain and the decoded signal in the frequency domain, And an inverse transform unit.

For the embodiments of the invention disclosed herein, specific structural and functional descriptions are set forth for the purpose of describing an embodiment of the invention only, and it is to be understood that the embodiments of the invention may be practiced in various forms, The present invention should not be construed as limited to the embodiments described in Figs.

The present invention is capable of various modifications and various forms, and specific embodiments are illustrated in the drawings and described in detail in the text. It should be understood, however, that the invention is not intended to be limited to the particular forms disclosed, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Similar reference numerals have been used for the components in describing each drawing.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are to be interpreted as either ideal or overly formal in the sense of the present application Do not.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used for the same constituent elements in the drawings and redundant explanations for the same constituent elements are omitted.

1 is a block diagram illustrating an audio / speech signal encoding apparatus according to an embodiment of the present invention.

1, an apparatus for encoding an audio / speech signal includes a frame determination unit 11, a domain conversion unit 12, a domain determination unit 13, a domain inversion unit 14, and an encoding unit 15 . The apparatus for encoding an audio / speech signal further includes a multiplexer 16.

The frame determination unit 11 variably determines the length of a frame as a processing unit of the input signal IN according to an attack position of the input signal IN by receiving the input signal IN. The input signal IN may be a pulse code modulation (PCM) signal obtained by modulating an analog speech signal or an audio signal with a digital signal, and the input signal IN may have an aperiodic attack.

Here, attack refers to the part that occurs when sound is divided into three stages of generation, continuation and extinction. For example, in an orchestra, it could be to start a single note like the beginning of a musical instrument. The attack time refers to the time during which the volume is peaking from immediately after the sound is generated, and the decay time corresponding to the time from the highest point to the midpoint of the volume. For example, when you hit the piano key 'land', the time until the 'ground' sound rises to the highest is called the attack time, and the time until the sound goes down from the peak is the decay time.

Here, a frame is a package of information transmitted in a unit of data communication or the like, and can be a unit of encoding and decoding. In particular, a frame may be a basic unit that can apply a fast fourier transform (FFT) to transform time domain data into the frequency domain. At this time, each frame can generate one frequency domain spectrum.

The conventional audio coding apparatus processes audio signals with a fixed frame length. For example, G.723.1 and G.729 of the International Telecommunication Union-Telecommunication Standardization Sector (ITU-T) are a typical coding technique. G.723.1 is a frame having a fixed length of 30 ms, G.729 is a frame having a fixed length of 10 ms . The AMR-NB (adaptive multi-rate-narrow band) coding apparatus operates with a fixed length frame of 20 ms. In the case of processing an audio signal with a frame having a fixed length, the audio signal is encoded without reflecting the characteristics of the input audio signal, for example, the position and intensity of the attack, Can be lowered.

Specifically, the frame determination unit 11 distinguishes the input signal IN as a stationary region and a transition region according to an attack position at which the sound of the input signal IN is generated. For example, the frame determination unit 11 may determine that the region where the attack exists in the input signal IN is a transition region, and the other region is a static region. The frame determining unit 11 can determine the length of the variable frame to be short according to the intensity of the attack of the input signal IN in the transition period and can determine the length of the variable frame in accordance with the static degree of the input signal IN, The length of the variable frame can be determined to be long depending on the range in which no attack exists.

More specifically, the frame determination unit 11 determines the length of the variable frame as the attack strength increases in the transition period in which the attack exists, . The resolution is usually used as an index indicating the accuracy of an image on a screen or the like. In the audio area, the time resolution indicates the temporal resolution of the audio signal, that is, the precision.

On the contrary, in the case of a static section in which there is no attack, the frame determining section 11 determines the length of the variable frame to be long according to the degree of staticness of the input signal IN, that is, The time resolution is limited by performing encoding for a longer interval, but it is possible to detect the change of the frequency and the input signal IN for a longer time, thereby increasing the frequency resolution. In the audio domain, the frequency resolution represents the resolution in the frequency direction, that is, the precision. This becomes clearer on the basis that time and frequency are inversely related.

As described above, the frame length is variably determined. In a region where the sound fluctuates as much as the transition period, the time resolution is increased, the frequency resolution is limited, and the frequency resolution is increased in the region where there is no sound variation , And coding is performed by restricting the time resolution, thereby improving the coding performance.

The frame determination unit 11 determines the length of the window when transforming the input signal IN in the time domain into the frequency domain according to the attack position of the input signal IN. Since the input signal IN is a PCM signal in the time domain, it is necessary to convert it into the frequency domain. Since the data to be processed in the discontinuous Fourier transform and the fast Fourier transform is a constant interval of the periodically repeated signal, when a conversion from the time domain to the frequency domain is performed, a certain period of the signal must be selected. In this manner, the window can be transformed from the time domain to the frequency domain by applying the window to the input signal IN of the time domain. Time and frequency are reciprocal. As the window width is narrow, the temporal resolution is improved. However, the frequency resolution is worse. If the window width is wide, the frequency resolution is improved, but the temporal resolution is worse. This is analogous to adjusting the length of the frame variably according to the position of the attack.

Further, the frame determination unit 11 can provide the encoding unit 15 with attack information such as an attack position and an intensity of the input signal IN.

The domain converter 12 converts the input signal IN into a frequency domain for each frame and divides the converted frequency domain signal by subband. Specifically, the domain conversion unit 12 receives the input signal IN and, based on the output of the frame determination unit 11, that is, based on the length of the frame determined by the frame determination unit 11, IN) is variably adjusted. The domain converter 12 divides the frequency domain converted signal into subbands and provides the divided signals to the domain determiner 13. [

For example, the input signal IN in the time domain is transformed into a frequency domain by a modified discrete cosine transform (MDCT), expressed as a real part, transformed into a frequency domain by a modified discrete sine transform (MDST) . Here, the signal converted by the MDCT and represented by the real part is used for encoding the input signal IN, and the signal converted by the MDST and expressed by the imaginary part is used for applying the psychoacoustic model.

The domain determination unit 13 determines whether the input signal IN is to be encoded in the frequency domain on a subband basis based on a frame whose length is determined to be different from each other according to the characteristics of the input signal IN, And decides whether to encode in the domain. Specifically, the domain determination unit 13 may use a spectral measure method such as a linear prediction coding gain, a spectrum change between linear prediction filters of adjacent frames, a spectral tilt of a frequency spectrum, A long term prediction estimation such as an energy measure method such as a signal energy size of each band, a signal energy change between bands, a predicted pitch delay, a predicted long term prediction gain, a long term prediction estimation method, and a voicing level determination method for discriminating voiced / unvoiced sounds.

The domain inverting unit 14 inversely converts the signal of the subband determined to be encoded in the time domain into the time domain based on the output of the domain determination unit 13, that is, the domain determination unit 13.

As described above, the frame length of the input signal IN is varied by the frame determination unit 11 and the domain determination unit 13, the input signal IN is divided into subbands in each frame, The domain to be encoded in the frequency band is determined. Thus, each frame and each frequency band are encoded in different domains.

The encoding unit 15 includes a frequency domain encoding unit 151 and a time domain encoding unit 152. The frequency domain encoding unit 151 receives the signal from the domain conversion unit 12 and encodes it in the frequency domain if it is determined to perform encoding in the frequency domain of the decision result of the domain determination unit 13. The time domain encoding unit 152 receives the signal from the domain inversion unit 14 and encodes it in the time domain when it is determined to perform the encoding in the time domain, as a result of determination by the domain determination unit 13.

In another embodiment, the encoding unit 15 includes a frequency domain encoding unit 151 and a time domain encoding unit 152, and the signals provided from the domain conversion unit 12 and the domain inversion unit 14 are And may be input to the frequency domain encoding unit. In this case, the time domain signal generated by the domain inversion unit 14 may be output to the time domain encoding unit. The encoding unit 15 may receive attack information such as an attack position and intensity of the input signal IN from the frame determination unit 11 and adaptively use the information to encode the input signal IN.

The multiplexing unit 16 receives the outputs of the frequency domain encoding unit 151 and the time domain encoding unit 151 and receives and multiplexes the results encoded in the frequency domain and the time domain, And generates a bit stream.

FIG. 2 is a graph illustrating a frame to be adjusted in an audio / speech signal encoding apparatus according to an embodiment of the present invention. Referring to FIG.

Referring to FIG. 2, the length of a frame of an input signal may be determined differently by a frame determination unit included in the audio / speech signal encoding apparatus of FIG. For example, the length of the first frame 21 is 15 ms, the length of the second and third frames 22 and 23 is 5 ms, the length of the fourth frame 24 is 10 ms, ) Is 5 ms. That is, the length of the first frame 21 is the longest, the length of the fourth frame 24 is long, and the lengths of the second, third and fifth frames 22, 23 and 25 are the shortest.

The second, third, and fifth frames 22, 23, 25 having the shortest frame length of 5 ms may be the transition period in which the attack is found. If an attack is found, the length of the frame can be shortened and the conversion window can be shortened to improve the temporal resolution. The first frame 21 having the longest frame length of 15 ms may be a static section in which no attack is found. If an attack is not found, the length of the frame can be adjusted to be long according to the degree of staticness, that is, the detection interval of the attack, and the frequency resolution can be improved by lengthening the conversion window.

FIG. 3 is a graph illustrating encoding domains for each frame and band in an audio / speech signal encoding apparatus according to an exemplary embodiment of the present invention. Referring to FIG.

Referring to FIGS. 2 and 3, the encoding domain of the input signal may be determined differently by the domain determination unit included in the audio / speech signal encoding apparatus of FIG. As described above, the domain determination unit can adaptively determine a domain that is advantageous for encoding according to the characteristics of the input signal for each frequency band. In FIG. 3, a white region is a frequency domain coding region and a black dot region is a time domain coding region.

For example, in the first frame 21, a frequency band 211 of 0 to 6 kHz is encoded in the time domain, and a frequency domain of 6 kHz to 10 kHz is encoded in the frequency domain. have. The frequency band 221 of 0 to 6 kHz of the second frame 22 is encoded in the time domain and the frequency domain of 6 kHz to 10 kHz is encoded in the frequency domain. The frequency band 231 of 0 to 6 kHz of the third frame 23 is coded in the time domain and the frequency domain of 6 kHz to 10 kHz is encoded in the frequency domain. In the case of the fourth frame 24, the frequency domain 240 of 0 to 10 kHz may all be determined to be encoded in the frequency domain. In the fifth frame 25, the frequency band 251 of 0 to 4 kHz may be encoded in the time domain and the frequency domain 252 of 4 kHz to 10 kHz may be encoded in the frequency domain.

In the conventional apparatus for encoding an audio / speech signal, the encoding domain is different for each frequency band in a frame having a predetermined length. However, the apparatus for encoding an audio / speech signal according to an exemplary embodiment of the present invention, And the encoding domain may be different for each frequency band in the frame. Thus, the length of the frame and the window type can be different according to the position and the intensity of the attack of the input signal, so that the time resolution and the frequency resolution can be improved.

4 is a block diagram illustrating an audio / speech signal decoding apparatus according to an embodiment of the present invention.

Referring to FIG. 4, the apparatus for decoding an audio / speech signal includes a demultiplexer 41, a determination unit 42, and a decoding unit 43. Further, the apparatus for decoding an audio / speech signal may further include a domain inverting unit 44.

The demultiplexer 41 receives and demultiplexes the bitstream, and outputs a result encoded in the frequency domain and a result encoded in the time domain.

The determination unit 42 determines a domain encoded according to the frame length and the frequency band of the demultiplexed signal based on the information obtained from the demultiplexed signal, and provides the determination result to the decoding unit 43. The encoding domain of the demultiplexed signal may be different for each frame length and frequency band.

The decoding unit 43 includes a frequency domain decoding unit 431 and a time domain decoding unit 432. More specifically, when the demultiplexed signal is encoded in the frequency domain, the frequency domain decoding unit 431 decodes the demultiplexed signal in the frequency domain. The time domain decoding unit 432 decodes the demultiplexed signal in the time domain when the demultiplexed signal is encoded in the time domain as a result of the determination by the determination unit 42. [

In another embodiment, the decoding unit 43 includes a frequency domain decoding unit 431 and a time domain decoding unit 432, and the demultiplexed signal can be input to the frequency domain decoding unit 431 once . In this case, when the demultiplexed signal is coded in the time domain as a result of the determination by the determination unit 42, the signal input to the frequency domain decoding unit 431 may be output to the time domain decoding unit 432 again.

The domain inverting unit 44 receives the output of the decoding unit 43, that is, receives the decoded signal, synthesizes the decoded signal in the time domain and the decoded signal in the frequency domain, and inversely converts the decoded signal into the time domain.

5 is a flowchart illustrating a method of encoding an audio / speech signal according to an embodiment of the present invention.

Referring to FIG. 5, in step 51, a length of a frame, which is a processing unit of an input signal, is variably determined according to an attack position of an input signal. Specifically, the input signal is classified into a static section or a transition section according to an attack position, and a length of a frame is determined differently in a static section and a transition section. For example, a long frame may be applied to the static section, and a short frame may be applied to the transition section depending on the attack strength.

In step 52, the input signal is transformed into the frequency domain for each frame, and the input signal is divided into subbands.

In step 53, it is determined whether the signals of the divided subbands are to be encoded in the frequency domain. In other words, it is determined whether the signals of the divided subbands are to be encoded in the frequency domain or in the time domain.

In step 54, the subband signal determined to be encoded in the frequency domain is encoded in the frequency domain.

In step 55, the signal of the subband determined to be encoded in the time domain is inversely transformed into the time domain and is encoded in the time domain.

6 is a flowchart illustrating a method of decoding an audio / speech signal according to an embodiment of the present invention.

Referring to FIG. 6, in step 61, the encoded domain of the signal and the subband is determined.

In step 62, the signal determined to be encoded in the time domain is decoded in the time domain, and the signal determined to be encoded in the frequency domain is decoded in the frequency domain.

In operation 63, the signal decoded in the time domain and the signal decoded in the frequency domain are synthesized and inverse transformed into the time domain.

It is needless to say that the present invention is not limited to the above-described embodiments, and can be modified by those skilled in the art within the scope of the present invention.

The present invention can also be embodied as computer-readable codes on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, flash memory, optical data storage, And the like. The computer readable recording medium may also be distributed over a networked computer system and stored and executed as computer readable code in a distributed manner.

As described above, according to the apparatus and method for encoding an audio / speech signal according to the present invention, encoding is performed according to a coding domain determined according to a length of a frame and a frequency band adjusted according to an attack position of an input signal, The compression efficiency and sound quality can be improved when the audio / speech signal is encoded by adjusting the frequency resolution.

According to the apparatus and method for decoding an audio / speech signal according to the present invention, decoding is performed adaptively according to a decoding domain determined according to a length of a frame and a frequency band, thereby adjusting a time resolution and a frequency resolution, The compression efficiency and the sound quality can be improved when the signal is decoded.

Claims (13)

delete delete delete delete delete delete delete delete delete delete Determining a domain encoded for each frame of the encoded signal; Generating a time domain reconstruction signal by performing time domain decoding on a signal determined to be encoded in a time domain, generating a time domain reconstruction signal by performing frequency domain decoding on a signal determined to be encoded in the frequency domain, And And providing the time domain restoration signal, which is generated by decoding in different domains, as a restoration signal of each frame, Wherein the long-term prediction and the fixed codebook are used in the time-domain decoding. A computer-readable recording medium having recorded thereon a program for executing the method of decoding an audio / speech signal according to claim 11. A determination unit for determining a domain encoded for each frame of the encoded signal; And The signal determined to be encoded in the time domain is time domain decoded to generate a time domain reconstructed signal, the signal determined to be encoded in the frequency domain is frequency domain decoded and then a time domain reconstructed signal is generated, And a decoding unit for providing the time domain restoration signal, which is generated by decoding in another domain, as a restoration signal of each frame, Wherein the long-term prediction and the fixed codebook are used in the time-domain decoding.

Images