KR20070091512A - Method and apparatus for error concealment of decoded audio signal - Google Patents

Abstract

A method and an apparatus for error concealment of decoded audio signals are provided to remove noise effectively by adaptively concealing an error according to a decoded bitstream and improve sound quality of audio by reflecting decoded information until an error is generated to the maximum. A method for error concealment of decoded audio signals comprises the following steps of: determining whether there is an error in the frame of a decoded bitstream(120); selecting a method for concealing the error in a frequency domain or a method for concealing the error in a time domain by a preset standard if it is determined that there is the error in the frame(145,150); and concealing the error according to the selected method(115).

Description

Method and apparatus for error concealment of decoded audio signal {Method and apparatus for error concealment of decoded audio signal}

1 is a flowchart illustrating an embodiment of a method of decoding an audio signal including a method of error concealment of a decoded audio signal according to the present invention.

2 is a block diagram illustrating an embodiment of an apparatus for decoding an audio signal including an error concealment apparatus of a decoded audio signal according to the present invention.

3 is a conceptual diagram illustrating a structure of a bitstream composed of a plurality of layers in a BSAC.

4A is a graph showing the spectrum energy of the previous frame in which no error exists.

4B is a graph showing the spectral energy of the current frame in which an error exists.

FIG. 5 is a conceptual diagram illustrating a transition using a method of concealing an error in a time domain.

6 is a conceptual diagram illustrating a method of recovering an error using a WSOLA method and frame interpolation in the time domain.

7 is a conceptual diagram illustrating an embodiment of applying the WSOLA scheme to n frames.

8 is a conceptual diagram illustrating an embodiment of applying frame interpolation.

<Brief description of the major symbols in the drawings>

200: bitstream input unit 205: header analyzer

210: decryption unit 225: determination unit

230: first selection unit 235: error position detection unit

240: second selector 253: frequency domain concealment

256: time domain concealment unit 260: inverse transform unit

The present invention relates to an apparatus for decoding audio data composed of a plurality of layers, such as BSAC. More particularly, the present invention relates to a concealment by recovering an error from a decoded bitstream. And a method and apparatus.

An error occurs in the process of transmitting the encoded audio signal through a wired or wireless network such as T-DMB or IP network. In this case, if the error is not handled properly, annoying distortion can occur, resulting in a significant degradation in sound quality.

Conventionally, in order to conceal the error of the audio signal, a muting method that reduces the effect of the error on the output by reducing the volume of sound in the error frame, the data of the previous frame (frame) in the error frame There may be a repetition method of copying and restoring, and a method of restoring a time domain sample of an error-prone frame by performing interpolation or extrainterpolation from a previous frame.

However, conventionally, in order to conceal an error existing in an audio signal, not only a portion in which an error occurs is restored, but all data corresponding to a unit frame including an error are restored by using another frame, thereby causing a large data loss. . As a result, the output sound quality is deteriorated.

An object of the present invention is to conceal an error by selecting from a frequency domain or a time domain in a method of concealing an error of a decoded bit stream, and an error exists in a frame when a method of concealing an error in a frequency domain is selected. The present invention provides a method and apparatus for error concealment of a restored audio signal that detects a position to conceal an error only in a portion corresponding to a later position.

According to an embodiment of the present invention, an error concealment method of a decoded audio signal includes determining whether an error exists in a frame of a decoded bitstream. If it is determined, the method of concealing the error in the frequency domain (concealment) or a method of concealing the error in the time domain according to a predetermined criterion and concealing the error according to the selected method Characterized in that it comprises a step.

The error concealment method of the decoded audio signal according to the present invention for achieving the above object is a method of concealing an error in the frequency domain or a time concealment of the error in a time domain by a predetermined criterion for the frame of the decoded bitstream having an error. Selecting a concealment scheme and concealing an error according to the selected scheme.

And a computer readable recording medium having recorded thereon a program for executing the above-described invention on a computer.

An error concealment apparatus of a decoded audio signal according to the present invention for achieving the above object is an error frame detection unit for detecting a frame of an error-decoded bitstream, in the frequency domain based on a predetermined reference to the detected frame And a concealment method selection unit for selecting among a method of concealing an error or a method of concealing an error in a time domain, and an error concealment unit for concealing an error according to the selected method.

The error concealment apparatus of the decoded audio signal according to the present invention for achieving the above object, the error concealment in the frequency domain or the manner of concealing the error in the frequency domain by a predetermined reference to the frame of the decoded bitstream with the error And a concealment method selection unit for selecting among concealment methods and an error concealment unit for concealing errors according to the selected method.

Hereinafter, a method and apparatus for error concealment of a decoded audio signal according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating an embodiment of a method of decoding an audio signal including a method of error concealment of a decoded audio signal according to the present invention.

First, a bitstream is received from a system of a higher level such as MPEG-4 system or MPEG TS (step 100).

In step 100, a header is extracted from the input bitstream, and information included in the header is analyzed (step 105).

After step 105, the bitstream received in step 100 is decoded (step 110).

In operation 115, an error concealment method of the decoded audio signal according to the present invention conceals an error existing in the bitstream decoded in operation 110 (operation 115).

In operation 110, it is determined whether an error exists in a current frame of the decoded bitstream (operation 120).

In step 120, whether there is an error in the current frame may be determined using the following three methods.

First, it is determined whether there is an error in the current frame by using a CRC or the like transmitted from the system and including information indicating whether the frame is in error.

Second, by comparing the length of the bit stream received from the encoder with the length of the bit stream decoded in step 110 determines whether there is an error in the current frame. If there is an error in the frame, a predetermined error occurs when the length of the decoded bitstream is compared with the length of the bitstream received from the encoder. In addition, when there is no error in a frame, the length of the bitstream received from the encoder is used for decoding. Therefore, when the length of the bitstream received from the encoder is the same as the length of the decoded bitstream, it is determined that there is no error in the bitstream of the current frame. If the length of the bitstream received from the encoder is different from the length of the decoded bitstream, it is determined that there is an error in the bitstream of the current frame.

Third, the number of bits of the unit frame included in the header of the bitstream is compared with the number of bits of the bitstream input in step 100 to determine whether there is an error in the current frame. For example, in the case of BSAC, if the frame_length information indicating the length of the unit frame provided in the header is compared with the number of bits of the bitstream input in step 100, and there is a difference more than a predetermined threshold value, there is an error in the current frame. To judge.

If it is determined in step 120 that there is an error in the current frame, the state of the current frame is analyzed to determine whether it is difficult to conceal the error in the frequency domain (step 125). In a codec using a bitstream provided with a plurality of layers such as BSAC or AAC illustrated in FIG. 3, it may be determined based on a window type whether it is difficult to conceal an error in a frequency domain. If the window type of the previous frame and the current frame is different due to the characteristics of the MDCT, the overlap and add operation is not performed even if there is no error, and thus a desired audio signal cannot be generated. Accordingly, in step 125, the state of the current frame is analyzed to determine whether it is difficult to conceal the error in the frequency domain by considering the state of the previous frame.

If it is determined in step 125 that it is not difficult to conceal the error in the frequency domain with respect to the current frame, a position where an error exists in the current frame is detected (step 130).

There are two methods for detecting a location where an error exists with respect to the current frame in operation 130.

First, the spectral energy of the current frame is compared with the spectral energy of the previous frame to detect a location where an error exists for the current frame. This is because when there is an error in a frame, the spectral energy of the current frame changes rapidly compared to the spectrum energy of the previous frame. For example, FIG. 4A is a graph showing the spectral energy of a previous frame in which there is no error, and FIG. 4B is a graph showing the spectral energy of the current frame in which there is an error. Comparing FIG. 4B with FIG. 4A, it can be seen that the spectral energy is drastically changed in the frequency domain where an error exists. Here, the spectral energy of the current frame and the spectral energy of the previous frame may be compared using the following equation.

[Equation 1]

Dt (f) = energy (f, t)-energy (f, t-1)

Where Dt (f) is the change in spectral energy between frames, energy (f, t) is the spectral energy of the current frame, and energy (f, t-1) is the spectral energy of the previous frame.

If Dt (f) is out of a predetermined range, it is determined that an error exists at the corresponding position.

In addition, the spectral energy of the current frequency domain and the spectral energy of the previous frequency domain are compared to detect a position where an error exists for the current frame. This is because if there is an error in the frame, the spectral energy of the current frequency domain is drastically changed compared to the spectrum energy of the previous frequency domain. This can be compared with the spectral energy of the current frequency domain and the spectral energy of the previous frequency domain by using the following equation.

[Equation 2]

Df (f) = energy (f, t)-energy (f-1, t)

Here, Df (f) is the change in spectral energy between frequency domains, energy (f, t) is the spectral energy of the current frequency domain, and energy (f, t-1) is the spectral energy of the previous frequency domain.

If Df (f) is out of a predetermined range, it is determined that an error exists at the corresponding position.

Using the above-described method, a layer corresponding to an error frequency band is derived.

Second, there is a method of detecting a location in which an error exists in a current frame by checking bits allocated to each layer of the decoded bitstream. In the case of a bitstream provided with a plurality of layers, information on the number of bits to be allocated to each layer is provided in a header. For example, in the case of BSAC, there is a help variable called layer_buf_offset indicating the number of bits to be allocated for each layer in the header. If there is an error in a given layer, excessively more or less bits are used in arithmetic decoding. Therefore, layers with too many or fewer bits are more likely to have errors in the current or previous layers.

In operation 130, it is determined whether the position detected in operation 130 is provided before the first threshold position. This is because when the position detected in operation 130 corresponds to data that plays an important role in decoding as in the base layer of FIG. 3, it is preferable to conceal an error in the time domain for the entire frame. However, unlike FIG. 1, the error may be recovered by using a method of concealing an error in the frequency domain for the entire frequency domain of the frame.

If it is determined in step 135 that the location is provided after the first critical location, it is determined whether the location detected in step 130 is provided after the second critical location (step 140). This is because the error does not need to be concealed when the position detected in operation 130 hardly affects the sound quality of the audio as in the last layer corresponding to the N layer of FIG. 3.

If it is determined in step 140 that it is provided before the second threshold position, an error is concealed in the frequency domain of the current frame (step 145). In step 145, a repetition method of copying and restoring data of a previous frame to a frame in which an error occurs is used to conceal an error. The data before the position detected in operation 130 is left as it is, and the data after the position detected in operation 130 is restored by copying data of a previous frame. Accordingly, when the bitstream is composed of a plurality of layers as in BSAC, and frequency bands are allocated to each layer, even if an error occurs in the middle of the bitstream, the previous layer in which the error occurs can be restored as it is. For example, when the position detected in operation 130 is 8 kHz, the layer corresponding to 8 kHz and the layer subsequent thereto are copied and restored to the layers of the corresponding previous frame. As a result, even in a frame in which an error occurs, the sound quality of the audio can be improved by reflecting the decoded information as much as possible before the error occurs.

If it is determined in step 125 that it is difficult to conceal the error in the frequency domain with respect to the current frame or in step 135, it is determined that the error is provided before the first threshold position (step 150).

In operation 150, the current frame is reconstructed by interpolation or extrapolation by retrieving a segment having the highest similarity from the previous frame, ignoring information on the current frame. When MDCT is applied, two PCM frames are lost because overlap and add operations must be performed. As a result, when an error occurs in one frame, two frames must be restored at all times. Therefore, in step 150, not only the current frame but also subsequent frames must be restored using interpolation or extrapolation. To this end, the information of the past frame is stored and the information of the past frame is used to conceal errors of the current frame. FIG. 5 illustrates a conceptual diagram to illustrate transitions using a method of concealing errors in the time domain. In addition, it is also possible to use synchronous overlap and add (recovery) by applying two consecutive PCM frames.

Referring to FIG. 6, when an error occurs in an n frame corresponding to the current frame, the 50% overlap window affects a (n + 1) frame corresponding to a subsequent frame. If the past signal is simply used repeatedly, a problem may occur in that the sound quality may be significantly degraded because the error occurs in succession of the last frame of the past, the first error frame and the second error frame. In order to solve this problem of continuous error, it is desirable to use the WSOLA method and the interpolation method simultaneously. The WSOLA method may be applied to n frames corresponding to the first frame, and frame interpolation may be used for (n + 1) frames corresponding to the second frame.

An embodiment of implementing the WSOLA method applied to n frames corresponding to the first frame is illustrated in FIG. 7. First, the output of the past time domain of the BSAC is buffered. The correlation is obtained by using the following equation in the past search range by using a block signal immediately before the current frame in which an error occurs.

[Equation 3]

Here, R (d) is a correlation, s (n) is an old_buff signal, and d is a sample position of a search range.

After finding the position of the found block having the highest correlation among the correlations obtained using Equation 3, the pound block and the next signal is copied and restored to the part where an error occurs. Here, when copying and restoring, windowing is performed, and a first-order function, a multiple-order function, and a sine function can be used for the portion of the pound block in which the overlap is made. The part can use a rectangular window. In addition, a window having a phase opposite to the window used in the pound block is also obtained in the block portion and overlapped with the portion of the pound block. Note that the window should be designed so that the sum of the two window values is 1 in the overlapping part. This same method is repeated until the end of the frame. Here, it may be divided and applied to at least one size according to the size of the frame. In addition, according to Equation 3, the correlation is obtained in every sample unit, but may be obtained in sample units corresponding to two or more integers in consideration of complexity.

In addition, in order to increase the precision, the correlation may be further obtained in units of one sample within a sample range corresponding to two or more integers around the sample having the highest correlation, and the position of the sample having the highest correlation may be obtained.

An embodiment of implementing the frame interpolation method applied to the (n + 1) frame corresponding to the second frame is shown in FIG. 8. Linear interpolation is applied between the old frame 800 and the next good frame 820. Due to the delay problem, the future signal is provided in units of frames, so that the window is modified for overlap and relaxed interpolation is applied. The window used in the relaxed interpolation method is shown in FIG. 8. As the window structure used herein, a rectangular window and a first-order function, a multiple-order function, or a sine function may be used. 8 illustrates one embodiment of a window using a linear function. In this case, it should be noted that when the previous window 800 and the subsequent window 820 overlap, the sum of the two latitude values should be 1. Referring to FIG. 8, a signal output by the relaxed interpolation method of overlapping and adding two windows having the same size as a frame is output by adding a block size (overlap size, OV_SIZE) to the signal of the frame size. The signal of the added block is then used to overlap and add the frame 820. In addition, the front part of the signal restored by the interpolation method is overlapped and added in the same manner as the method described with reference to FIG. 7 with the previous frame 800.

In operation 120, it is determined that there is no error in the current frame, or in operation 140, after the second critical position is determined, or after operation 145, an inverse transform is performed to restore the final audio output (operation 155). ).

2 is a block diagram illustrating an apparatus for decoding an audio signal including an error concealment apparatus of a decoded audio signal according to the present invention, wherein the apparatus for decoding the audio signal includes a bitstream input unit 200. , A header analyzer 205, a decoder 210, an error concealment device 220 of the decoded audio signal, and an inverse transformer 260.

The bitstream input unit 200 receives a bitstream from an upper level system such as MPEG-4 system or MPEG TS through the input terminal IN.

The header analyzer 205 extracts a header from the bitstream input from the bitstream input unit 200 and analyzes the information included in the header.

The decoder 210 decodes the bitstream received from the bitstream input unit 200.

The error concealment apparatus 220 of the decoded audio signal conceals an error existing in the bitstream decoded by the decoder 210. Here, the error concealment apparatus 220 of the decoded audio signal includes a determination unit 225, a first selection unit 230, an error position detector 235, a second selection unit 240, and an error concealment unit 250. It is made to include.

The determiner 225 determines whether there is an error in the current frame of the bitstream decoded by the decoder 210. The determination unit 225 may determine whether there is an error in the current frame by using the following three methods.

First, it is determined whether there is an error in the current frame by using a CRC or the like transmitted from the system and including a cannon indicating whether the frame is in error.

Second, by comparing the length of the bit stream received from the encoder with the length of the bit stream decoded by the decoder 210 determines whether there is an error in the current frame. If there is an error in the frame, a predetermined error occurs when the length of the decoded bitstream is compared with the length of the bitstream received from the encoder. In addition, when there is no error in a frame, the length of the bitstream received from the encoder is used for decoding. Therefore, when the length of the bitstream received from the encoder is the same as the length of the decoded bitstream, it is determined that there is no error in the bitstream of the current frame. If the length of the bitstream received from the encoder is different from the length of the decoded bitstream, it is determined that there is an error in the bitstream of the current frame.

Third, by comparing the number of bits of the unit frame included in the header of the bitstream with the number of bits of the bitstream received from the bitstream input unit 200 determines whether there is an error in the current frame. For example, in case of BSAC, if the frame_length information indicating the length of the unit frame provided in the header is compared with the number of bits of the bitstream input from the bitstream input unit 200 and there is a difference over a predetermined threshold value, an error occurs in the current frame. Judging that there is.

When the determination unit 225 determines that there is an error in the current frame, the first selector 230 analyzes the state of the current frame to determine whether it is difficult to conceal the error in the frequency domain, thereby concealing the error in the frequency domain. Or choose to conceal errors in the time domain. In a codec using a bitstream provided with a plurality of layers such as BSAC or AAC illustrated in FIG. 3, it may be determined based on a window type whether it is difficult to conceal an error in a frequency domain. If the window type of the previous frame and the current frame is different due to the characteristics of the MDCT, the overlap and add operation is not performed even if there is no error, and thus a desired audio signal cannot be generated. Accordingly, the first selector 230 analyzes the state of the current frame to determine whether it is difficult to conceal the error in the frequency domain by considering the state of the previous frame. Therefore, when it is determined that it is difficult to conceal the error in the frequency domain with respect to the current frame, the first selector 230 selects a method of concealing the error in the time domain, and when it is determined that it is not difficult to conceal the error in the frequency domain with respect to the current frame. Choose how to conceal errors in the frequency domain.

If a method of concealing an error in the frequency domain is selected by the first selector 230, the error location detector 235 detects a location where an error exists in the current frame. The error location detector 235 detects a location in which an error exists with respect to the current frame.

First, the spectral energy of the current frame is compared with the spectral energy of the previous frame to detect a location where an error exists for the current frame. This is because when there is an error in a frame, the spectral energy of the current frame changes rapidly compared to the spectrum energy of the previous frame. For example, FIG. 4A is a graph showing the spectral energy of a previous frame in which there is no error, and FIG. 4B is a graph showing the spectral energy of the current frame in which there is an error. Comparing FIG. 4B with FIG. 4A, it can be seen that the spectral energy is drastically changed in the frequency domain where an error exists. Here, the spectral energy of the current frame and the spectral energy of the previous frame may be compared using the following equation.

[Equation 4]

Dt (f) = energy (f, t)-energy (f, t-1)

Here, Dt (f) is the change in spectral energy between frames, energy (f, t) is the spectral energy of the current frame, and energy (f, t-1) is the spectral energy of the previous frame.

If Dt (f) is out of a predetermined range, it is determined that an error exists at the corresponding position.

In addition, the spectral energy of the current frequency domain and the spectral energy of the previous frequency domain are compared to detect a position where an error exists for the current frame. This is because if there is an error in the frame, the spectral energy of the current frequency domain is drastically changed compared to the spectrum energy of the previous frequency domain. This can be compared with the spectral energy of the current frequency domain and the spectral energy of the previous frequency domain by using the following equation.

[Equation 5]

Df (f) = energy (f, t)-energy (f-1, t)

Here, Df (f) is the change in spectral energy between frequency domains, energy (f, t) is the spectral energy of the current frequency domain, and energy (f, t-1) is the spectral energy of the previous frequency domain.

If Df (f) is out of a predetermined range, it is determined that an error exists at the corresponding position.

Using the above-described method, a layer corresponding to an error frequency band is derived.

Second, there is a method of detecting a location in which an error exists in a current frame by checking bits allocated to each layer of the decoded bitstream. In the case of a bitstream provided with a plurality of layers, information on the number of bits to be allocated to each layer is provided in a header. For example, in the case of BSAC, there is a help variable called layer_buf_offset indicating the number of bits to be allocated for each layer in the header. If there is an error in a given layer, excessively more or less bits are used in arithmetic decoding. Therefore, layers with too many or fewer bits are more likely to have errors in the current or previous layers.

The second selector 240 selects a method of concealing an error in the frequency domain or a method of concealing the error in the time domain based on the location of the error detected by the error location detector 235.

The second selector 240 may select a method of concealing an error in the time domain when the position detected by the error position detector 235 is provided before the first threshold position. This is because when the position detected by the error position detector 235 corresponds to data that plays an important role in decoding as in the base layer of FIG. 3, it is preferable to conceal an error in the time domain for the entire frame. However, unlike shown in FIG. 1, the second selector 240 may select a method of concealing an error in a frequency domain that recovers the error by using a method of concealing an error in the frequency domain for the entire frequency domain of the frame. have.

When the position detected by the error position detector 235 is provided after the second threshold position, the second selector 240 outputs the inverse transform unit 260 without concealing the error. This is because, if the position detected by the error position detector 235 is the last layer corresponding to the N layer of FIG. 3, the error does not have to conceal the error when the error hardly affects the sound quality of the audio.

The second selector 240 selects a method of concealing an error in the frequency domain when the position detected by the error position detector 235 is provided between the first threshold position and the second threshold position.

The error concealment unit 250 recovers and conceals an error of the current frame. Here, the error concealment part 250 includes a frequency domain concealment part 253 and a time domain concealment part 256.

The frequency domain concealment unit 253 conceals an error of the current frame in the frequency domain. Here, the frequency domain concealment unit 235 preferably uses an iterative method of copying and restoring data of a previous frame to a frame in which an error occurs in concealing an error of a current frame. The data before the position detected by the error position detection unit 235 is left as it is, and the data after the position detected by the error position detection unit 235 copies and restores the data of the previous frame. Accordingly, when the bitstream is composed of a plurality of layers as in BSAC, and frequency bands are allocated to each layer, even if an error occurs in the middle of the bitstream, the previous layer in which the error occurs can be restored as it is. For example, when the position detected in operation 130 is 8 kHz, the layer corresponding to 8 kHz and the layer subsequent thereto are copied and restored to the layers of the corresponding previous frame. As a result, even in a frame in which an error occurs, the sound quality of the audio can be improved by reflecting the decoded information as much as possible before the error occurs.

The time domain concealment unit 256 conceals an error of the current frame in the time domain.

In concealing the error in the time domain concealment unit 256, the information on the current frame is ignored and the part with the highest similarity is retrieved from the previous frame to restore the current frame by interpolation or extrapolation. When MDCT is applied, two PCM frames are lost because overlap and add operations must be performed. As a result, when an error occurs in one frame, two frames must be restored at all times. Therefore, in the time domain concealment unit 256, not only the current frame but also subsequent frames must be restored using interpolation or extrapolation. To this end, by storing information of the past frame and using the information of the past frame, the error of the current frame is concealed. FIG. 5 illustrates a conceptual diagram to illustrate a transformation using a method of concealing errors in the time domain. It is also possible to use synchronous overlap and add, which applies and recovers two consecutive PCM frames.

Referring to FIG. 6, when an error occurs in an n frame corresponding to the current frame, the 50% overlap window affects a (n + 1) frame corresponding to a subsequent frame. If the past signal is simply used repeatedly, a problem may occur in that the sound quality may be significantly degraded because the error occurs in succession of the last frame of the past, the first error frame and the second error frame. In order to solve this problem of continuous error, it is desirable to use the WSOLA method and the interpolation method simultaneously. WSOLA can be applied to n frames corresponding to the first frame, and frame interpolation can be used for (n + 1) frames corresponding to the second frame.

An embodiment of implementing the WSOLA method applied to n frames corresponding to the first frame is illustrated in FIG. 7. First, the output of the past time domain of the BSAC is buffered. The correlation is obtained by using the following equation in the past search range by using a block signal immediately before the current frame in which an error occurs.

[Equation 6]

Here, R (d) is a correlation, s (n) is an old_buff signal, and d is a sample position of a search range.

Using Equation 6, after finding the position of the pound block having the highest correlation among the correlations obtained, the pound block and the next signal are copied and restored to the part where an error occurs. Here, when copying and restoring, windowing is performed. A first-order function, a multiple-order function, and a sine function can be used for the part of the pound block that is overlapped, and a rectangular window is used for the part of the pound block that does not overlap. Can be. In addition, a window having a phase opposite to the window used in the pound block is also obtained in the block portion and overlapped with the portion of the pound block. Note that the window should be designed so that the sum of the two window values is 1 in the overlapping part. This same method is repeated until the end of the frame. Here, it may be divided and applied to at least one size according to the size of the frame. In addition, according to Equation 3, the correlation is obtained in every sample unit, but may be obtained in sample units corresponding to two or more integers in consideration of complexity.

In addition, in order to increase the precision, the correlation may be further obtained in units of one sample within a sample range corresponding to two or more integers around the sample having the highest correlation, and the position of the sample having the highest correlation may be obtained.

An embodiment of implementing the frame interpolation method applied to the (n + 1) frame corresponding to the second frame is shown in FIG. 8. Linear interpolation is applied between the previous frame 800 and the subsequent frame 820 without errors. Because of the delay, the future signal is frame-by-frame, so the window is modified for overlap and relaxed interpolation is applied. The window used in the relaxed interpolation method is shown in FIG. 8. As the window structure used herein, a rectangular window and a first-order function, a multiple-order function, or a sine function may be used. 8 illustrates one embodiment of a window using a linear function. In this case, it should be noted that when the previous window 800 and the subsequent window 820 overlap, the sum of the two upper window values should be 1. Referring to FIG. 8, a signal output by the relaxed interpolation method of overlapping and adding two windows having the same size as a frame is output by adding a block of an overlap size (OV_SIZE) to a signal having a frame size. The signal of the added block is then used to overlap and add the frame 820. In addition, the front part of the signal restored by the interpolation method is overlapped and added in the same manner as the method described with reference to FIG. 7 with the previous frame 800.

The inverse transform unit 260 restores the final audio output by performing inverse transformation and outputs it through the output terminal OUT.

The present invention can be embodied as code that can be read by a computer (including all devices having an information processing function) in a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording devices include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like.

Although described with reference to the embodiment shown in the drawings to aid in understanding of the present invention, this is merely exemplary, those skilled in the art that various modifications and equivalent other embodiments are possible from this. Will understand. Therefore, the true technical protection scope of the present invention will be defined by the appended claims.

According to a method and apparatus for concealing an error of a decoded audio signal according to the present invention, a method of concealing an error of a decoded bit stream conceals an error by selecting from a frequency domain or a time domain and conceals an error in a frequency domain. If this is selected, the position where the error exists in the frame is detected and only the portion corresponding to the subsequent position conceals the error.

By doing this, noise can be effectively concealed according to the decoded bitstream, thereby effectively eliminating noise. In addition, even in the case of an error frame, since the information decoded up to the maximum until the error occurs, the sound quality of the audio can be improved.

Claims (37)

Determining whether there is an error in a frame of the decoded bitstream; If it is determined that there is an error in the frame, selecting one of a method of concealing an error in a frequency domain or a method of concealing an error in a time domain according to a predetermined criterion; And Concealing the error according to the selected scheme. The method of claim 1, wherein the selecting of the preset criteria comprises: Determining whether it is difficult to conceal errors in the frequency domain for the frame; And And a method of concealing an error in a frequency domain or a method of concealing an error in a time domain according to the determined result. The method of claim 2, wherein the determining step And determining whether or not it is difficult to conceal errors in the frequency domain with respect to the frame based on a window type. The method of claim 2, wherein the selecting of the preset criteria comprises: Detecting a location where an error exists in the frame; And And selecting a method of concealing an error in a frequency domain or a method of concealing an error in a time domain by using the detected position. The method of claim 4, wherein the detecting step The method for concealing an error of a decoded audio signal, characterized in that is performed only when a method of concealing an error in a frequency domain is selected in the step of selecting according to the determined result. The method of claim 4, wherein the detecting step And detecting a position where an error exists in the frame by comparing the spectrum energy of the frame and the spectral energy of a previous frame. The method of claim 4, wherein the detecting step And detecting the position of the error in the frame by comparing the spectral energy of the frequency domain with the spectral energy of the previous frequency domain. The method of claim 4, wherein the detecting step And detecting a location where an error exists in the layer by inspecting a bit allocated to each layer of the decoded bit stream. The method of claim 4, wherein the selecting according to the detected position comprises: And selecting a method of concealing an error in a time domain when the detected position is provided before the first threshold position. The method of claim 4, wherein the selecting according to the detected position comprises: And a method of concealing an error in a frequency domain when the detected position is within a preset range. The method of claim 4, wherein the selecting according to the detected position comprises: And if the detected position is provided after a second threshold position, error concealment of the decoded bit stream. The method of claim 4, wherein the concealing step And if a method of concealing errors in a frequency domain is selected, restoring a frequency band corresponding to the detected position to a signal corresponding to a frequency band of a previous frame. The method of claim 4, wherein the concealing step If a method of concealing an error in the frequency domain is selected, restoring the layer including the detected position and subsequent layers to the layers of the corresponding previous frame. The method of claim 1, wherein the concealing step If the method of concealing an error in the time domain is selected, concealing the error using interpolation or extrapolation. The method of claim 1, wherein the concealing step If a method of concealing an error in the time domain is selected, concealing the error for the frame and the next frame. The method of claim 15, wherein the concealing step Concealing an error in a WSOLA method for the frame and concealing the interpolation method for the next frame. The method of claim 1, wherein the determining And comparing the length of the received bitstream with the length of the decoded bitstream to determine whether there is an error in the frame of the decoded bitstream. Selecting a method of concealing an error in a frequency domain or a method of concealing an error in a time domain by a predetermined criterion for a frame of an error-decoded bitstream; And Concealing the error according to the selected scheme. A computer-readable recording medium having recorded thereon a program for executing the invention according to any one of claims 1 to 18 on a computer. An error frame detector for detecting a frame of an error-decoded bitstream; A concealment method selection unit for selecting a method of concealing an error in a frequency domain or a method of concealing an error in a time domain based on a predetermined reference with respect to the detected frame; And And an error concealment unit for concealing an error according to the selected scheme. The method of claim 20, wherein the concealment method selection unit A determination unit that determines whether it is difficult to conceal an error in the frequency domain with respect to the frame; And And a first selector configured to select a method of concealing an error in a frequency domain or a method of concealing an error in a time domain according to the determined result. The method of claim 21, wherein the determination unit And determining whether or not it is difficult to conceal an error in the frequency domain with respect to the frame based on a window type. The method of claim 21, wherein the concealment method selection unit An error location detector for detecting a location where an error exists in the frame; And And a second selector configured to select a method of concealing an error in a frequency domain or a method of concealing an error in a time domain by using the detected position. The method of claim 23, wherein the error position detection unit And detecting a position where an error exists in the frame only when a method of concealing an error in a frequency domain is selected by the first selector. The method of claim 23, wherein the error position detection unit And detecting the position of the error in the frame by comparing the spectral energy of the frame with the spectral energy of the previous frame. The method of claim 23, wherein the error position detection unit And detecting the position of the error in the frame by comparing the spectral energy of the frequency domain with the spectral energy of the previous frequency domain. The method of claim 23, wherein the error position detection unit And detecting a location where an error exists in the frame by checking a bit allocated to each layer of the decoded bit stream. The method of claim 23, wherein the second selection unit And selecting a method of concealing an error in a time domain when the detected position is provided before a first threshold position. The method of claim 23, wherein the second selection unit And a method of concealing an error in a frequency domain when the detected position is included in a preset range. The method of claim 23, wherein the second selection unit And if the detected position is provided after a second threshold position, the error concealment apparatus of the decoded audio signal, wherein the error of the decoded bitstream is not concealed. The method of claim 23, wherein the error concealment is If the method of concealing the error in the frequency domain is selected, error concealment apparatus of the decoded audio signal, characterized in that for recovering the frequency band corresponding to the detected position to the signal corresponding to the frequency band of the previous frame. The method of claim 23, wherein the error concealment is And if a method of concealing an error in the frequency domain is selected, restoring the layer including the detected position and subsequent layers to the layers of the corresponding previous frame. The method of claim 20, wherein the error concealment unit If the method of concealing the error in the time domain is selected, error concealment apparatus of the decoded audio signal, characterized in that concealment of the error using interpolation or extrapolation. The method of claim 20, wherein the error concealment unit And if a method of concealing an error in the time domain is selected, concealing the error for the frame and the next frame. 35. The method of claim 34, wherein the error concealment is Error concealment apparatus for the decoded audio signal according to the WSOLA method for the frame, and concealed by the interpolation method for the next frame. The method of claim 20, wherein the error frame detection unit And a frame of the decoded bitstream in which an error is detected by comparing the length of the received bitstream with the length of the decoded bitstream. A concealment method selection unit for selecting a method of concealing an error in a frequency domain or a method of concealing an error in a time domain based on a predetermined criterion for a frame of an error-decoded bitstream; And And an error concealment unit for concealing an error according to the selected scheme.

Images