RU2575680C2 - Device and method for encoding/decoding for high-frequency band extension - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: encoding device can reduce input signal sampling, perform basic encoding of the input signal with reduced sampling, perform frequency conversion of the input signal and perform encoding with bandwidth extension using the basic signal of the input signal in the frequency domain.

EFFECT: high accuracy of encoding and decoding a signal in high-frequency band extension modes.

20 cl, 38 dwg

Description

Technical field

One or more aspects of the present invention relates to a method and apparatus for encoding and decoding an audio signal, such as a speech signal or a music signal, and more particularly, to a method and apparatus for encoding and decoding a signal corresponding to a high frequency range of an audio signal.

State of the art

A signal corresponding to the high frequency range is less sensitive to the exact frequency structure than a signal corresponding to the low frequency range. Thus, when the encoding efficiency is improved to remove limitations on the bits available for encoding an audio signal, a large number of bits are assigned to a signal corresponding to a low frequency range, and a relatively small number of bits are assigned to a signal corresponding to a high frequency range.

A technology using the above method is spectral band copying (SBR). In SBR, coding efficiency is improved by expressing the high-frequency signal with the envelope and synthesizing the envelope during the decoding process. SBR is based on the characteristics of human hearing, having a relatively low resolution in relation to the high-frequency signal.

DETAILED DESCRIPTION OF THE INVENTION

Technical problem

The present invention provides methods for expanding a high frequency range based on a copy of spectral bands (SBR).

Technical solution

In accordance with an aspect of the present invention, there is provided an encoding apparatus including: a down-sampler for down-sampling an input signal; a basic encoder for performing basic coding of an input signal with downsampling; a frequency converter for performing frequency conversion on the input signal; and an extension encoder for performing band extension coding using the main signal of the input signal in the frequency domain.

In accordance with another aspect of the present invention, there is provided an encoding apparatus including: a down-sampler for down-sampling an input signal; a basic encoder for performing basic coding of an input signal with downsampling; a frequency converter for performing frequency conversion on the input signal; and an extension encoder for performing band extension coding using the characteristics of the input signal and the main signal of the input signal in the frequency domain.

In accordance with another aspect of the present invention, there is provided an encoding apparatus including an encoding mode selector for selecting an encoding mode for performing band expansion based on an input signal in a frequency domain and an input signal in a time domain and an extension encoder for performing extension encoding frequency bands using the input signal in the frequency domain and the encoding mode.

An extension encoder may include a main signal generator for generating a main signal of an input signal in a frequency domain from a frequency spectrum of an input signal in a frequency domain; a coefficient estimation module for estimating an energy regulation coefficient using a main signal; an energy extraction module for extracting energy from an input signal in a frequency domain; an energy controller for controlling the extracted energy using an energy control coefficient; and an energy quantizer for quantizing the regulated energy.

An extension encoder may include a main signal generator for generating a main signal of an input signal in a frequency domain using a frequency spectrum of an input signal in a frequency domain; a coefficient estimating module for estimating an energy regulation coefficient based on using characteristics of an input signal and a main signal; an energy extraction module for extracting energy from an input signal in a frequency domain; an energy controller for controlling the extracted energy using an energy control coefficient; and an energy quantizer for quantizing the regulated energy.

An extension encoder may include an energy extraction module for extracting energy from an input signal in a frequency domain based on a coding mode; an energy controller for controlling the extracted energy using an energy regulation coefficient based on a coding mode; and an energy quantizer for quantizing the regulated energy based on a coding mode.

The main signal generator may include an artificial signal generator for generating an artificial signal corresponding to the high frequency range of the high frequency region by copying and collapsing the low frequency range of the low frequency region of the input signal in the frequency domain; an envelope estimator for estimating an envelope of the main signal using a window; and an envelope application unit for applying the estimated envelope to the artificial signal.

The coefficient estimation module may include a first tonality calculator for calculating a tonality of a high frequency range of a high frequency region of an input signal in a frequency region; a second tonality calculator for calculating the tonality of the main signal and a coefficient calculator for calculating the energy control coefficient using the tonality of the high frequency range of the high frequency region of the input signal in the frequency domain and the tonality of the main signal.

In accordance with another aspect of the present invention, there is provided an encoding apparatus including a signal classification unit for determining an encoding mode of an input signal based on characteristics of an input signal; Code Excited Linear Prediction Encoder (CELP) for performing CELP coding of a low frequency signal of an input signal when an encoding mode of an input signal is defined as a CELP encoding mode; time domain (TD) encoder with extension for performing extension coding of a high frequency signal of an input signal when a low frequency signal of an input signal is encoded with CELP; a frequency converter for performing frequency conversion on an input signal when an encoding mode of an input signal is defined as a frequency domain (FD) mode and an encoder in FD for performing FD encoding of a converted input signal.

The encoder in the FD may include a normalization encoder for extracting energy from the converted input signal for each frequency band and quantizing the extracted energy; a factorial pulse encoder for performing factorial pulse encoding (FPC) of the value obtained by scaling the converted input signal using the quantized normalization value and an additional noise information generator to generate additional noise information in accordance with the FPC, and the converted input signal input to the encoder FD may be a transition frame.

The encoder in the FD may include a normalization encoder for extracting energy from the converted input signal for each frequency band and quantizing the extracted energy; a factorial pulse encoder for performing factorial pulse encoding (FPC) of a value obtained by scaling a converted input signal using a quantized normalization value; an additional noise information generator for generating additional noise information in accordance with the implementation of the FPC; and an encoder in FD with extension for performing extension coding of the high frequency signal of the converted input signal, and the converted input signal input to the FD encoder may be a constant frame.

An encoder in an FD with extension can perform energy quantization using the same codebook at different bit rates.

The stream of binary signals in accordance with the result of the FD encoding of the converted input signal may include information about the mode of the previous frame.

In accordance with another aspect of the present invention, there is provided an encoding apparatus including a signal classification unit for determining an encoding mode of an input signal based on characteristics of an input signal; linear prediction coefficient encoder (LPC) for extracting LPC from a low frequency input signal and quantizing LPC; Code Excited Linear Prediction Encoder (CELP) for performing CELP encoding of the LPC excitation signal of the low frequency signal of the input signal extracted using the LPC when the encoding mode of the input signal is defined as the CELP encoding mode; time-domain encoder (TD) with extension for performing extension coding of a high-frequency signal of an input signal when an LPC drive signal is encoded with CELP; an audio encoder for performing audio coding of the LPC drive signal when the encoding mode of the input signal is determined to be an audio mode; and an encoder in FD with extension for performing extension coding of the high frequency signal of the input signal when the LPC drive signal is subjected to audio coding.

An encoder in an FD with extension can perform energy quantization using the same codebook at different bit rates.

In accordance with another aspect of the present invention, there is provided a decoding apparatus including a base decoder for performing basic decoding of a base-coding input signal included in a binary signal stream; an upsampler for upsampling an input signal decoded by basic decoding; a frequency converter for performing frequency conversion on an input signal with an increased sampling frequency and an extension decoder for performing wideband decoding using the energy of the input signal included in the binary signal stream and the input signal in the frequency domain.

The expansion decoder may include an inverse quantizer for inverse quantization of the energy of the input signal; a main signal generator for generating a main signal using an input signal in a frequency domain; gain calculator for calculating the gain to be applied to the main signal using the inverse quantized energy and the energy of the main signal; and a gain applying unit for applying gain to each of the frequency bands.

The main signal generator may include an artificial signal generator for generating an artificial signal corresponding to the high frequency range by copying and folding the low frequency range of the input signal in the frequency domain; envelope estimation module for estimating the envelope of the main signal using a window included in the binary signal stream; and an envelope application unit for applying the estimated envelope to the artificial signal.

In accordance with another aspect of the present invention, there is provided a decoding apparatus including: a mode information checking unit for checking mode information of each of frames included in a binary signal stream; Code Excited Linear Prediction (CELP) decoder for performing CELP decoding of a CELP encoded frame based on the verification result; a time domain (TD) decoder with extension for generating a decoded highband signal using at least one of a result of decoding with a CELP and a low frequency excitation signal; a frequency domain decoder (FD) for performing FD decoding of an FD-encoded frame based on the test result; and an inverse frequency converter for performing inverse frequency conversion on the result of performing FD decoding.

The decoder in the FD may include a normalization decoder for performing normalization decoding based on normalization information included in the binary signal stream; a factorial pulse coding (FPC) decoder for performing FPC decoding based on factorial pulse coding information included in a binary signal stream; and a noise filling execution unit for performing noise filling of an FPC decoding execution result.

The decoder in the FD may include a normalization decoder for performing normalization decoding based on normalization information included in the binary signal stream; a factorial pulse coding (FPC) decoder for performing FPC decoding based on factorial pulse coding information included in a binary signal stream; a noise filling execution unit for performing noise filling of an FPC decoding execution result; and a decoder in FD with high frequency extension for performing high frequency extension decoding based on the result of performing FPC decoding and the result of performing noise filling.

The decoder in the FD may further include an encoder in the FD with low frequency extension to perform encoding with the extension of the results of FPC decoding and noise filling when the upper band value of the frequency band performing FPC decoding is less than the upper band value of the base signal band.

A high frequency extension FD decoder can perform energy quantization backwards by sharing the same codebook at different bit rates.

A decoder in FD may perform FD decoding of a frame encoded in FD based on information about the previous frame mode included in the binary signal stream.

In accordance with another aspect of the present invention, there is provided a decoding apparatus including: a mode information checking unit for checking mode information of each of frames included in a binary signal stream; a linear prediction coefficient (LPC) decoder for decoding LPC frames included in the binary signal stream; Code Excited Linear Prediction (CELP) decoder for performing CELP decoding of a CELP encoded frame based on the verification result; a time domain (TD) decoder with extension for generating a decoded highband signal using at least one of a result of decoding with a CELP and a low frequency excitation signal; an audio decoder for performing audio decoding of the encoded audio frame based on the verification result; and a frequency domain (FD) decoder with extension for performing extension decoding using the result of performing audio decoding.

An extension decoder in FD can perform energy quantization backwards by sharing the same codebook at different bit rates.

In accordance with another aspect of the present invention, there is provided an encoding method including down-sampling an input signal; performing basic coding of the input signal with downsampling; performing frequency conversion on the input signal and performing coding with the extension of the frequency band using the main signal of the input signal in the frequency domain.

In accordance with another aspect of the present invention, there is provided an encoding method including down-sampling an input signal; performing basic coding of the input signal with downsampling; performing frequency conversion of the input signal and performing coding with the extension of the frequency band using the characteristics of the input signal and the main signal of the input signal in the frequency domain.

In accordance with another aspect of the present invention, there is provided an encoding method including selecting an encoding mode for performing band expansion based on the use of an input signal in a frequency domain and an input signal in a time domain, and performing band expansion encoding using an input signal in frequency domain and coding mode.

In accordance with another aspect of the present invention, a decoding method is provided, which includes performing basic decoding of an input signal encoded by a base code included in a binary stream; upsampling of the input signal decoded by the base code; performing frequency conversion of the input signal with an increased sampling frequency and performing decoding with the extension of the frequency band using the input signal in the frequency domain and the energy of the input signal included in the binary signal stream.

In accordance with another aspect of the present invention, there is provided an encoding method including determining an encoding mode of an input signal based on characteristics of the input signal; performing code-excited linear prediction coding (CELP) coding of the low frequency signal of the input signal when the coding mode of the input signal is determined to be a CELP coding mode; performing coding in the time domain (TD) with the extension of the high-frequency signal of the input signal when the low-frequency signal of the input signal is encoded with CELP; performing frequency conversion of the input signal when the encoding mode of the input signal is determined as the frequency domain mode (FD); and performing FD encoding of the converted input signal.

Performing FD coding may include performing energy quantization by sharing a single codebook at different bit rates.

The stream of binary signals in accordance with the result of the FD encoding of the converted input signal may include information about the mode of the previous frame.

In accordance with another aspect of the present invention, there is provided an encoding method including determining an encoding mode of an input signal based on characteristics of the input signal; extracting linear prediction coefficients (LPC) from the low-frequency signal of the input signal and quantizing the LPC; performing code-excited linear prediction coding (CELP) of the LPC drive signal of the low-frequency signal of the input signal extracted using the LPC when the coding mode of the input signal is determined to be the CELP coding mode; performing coding in the time domain (TD) with the extension of the high-frequency signal of the input signal when the LPC excitation signal is subjected to coding with CELP; performing audio coding of the LPC drive signal when the coding mode of the input signal is determined to be an audio coding mode; and performing frequency domain coding (FD) with the extension of the high frequency signal of the input signal when the LPC drive signal is subjected to audio coding.

Performing FD coding with the extension may include performing energy quantization by sharing a single codebook at different bit rates.

In accordance with another aspect of the present invention, a decoding method is provided, including: checking mode information of each of the frames included in a binary signal stream; performing code-excited linear prediction (CELP) decoding encoded with a CELP frame based on the verification result; generating a decoded high-frequency band signal using at least one of a result of decoding with CELP and a low-frequency signal drive signal; performing decoding in the frequency domain (FD) of the frame encoded in FD based on the verification result; and performing inverse frequency conversion of the result of the FD decoding.

Performing FD decoding may include performing energy inverse quantization by sharing a single codebook at different bit rates.

Performing FD decoding may include performing FD decoding of an FD encoded frame based on information about a previous frame mode included in the binary signal stream.

In accordance with another aspect of the present invention, a decoding method is provided, including: checking mode information of each of the frames included in a binary signal stream; performing decoding of linear prediction coefficients (LPC) of frames included in the binary signal stream; performing code-excited linear prediction (CELP) decoding of a frame encoded with CELP based on the verification result; generating a decoded signal of the high-frequency signal using at least one of the result of decoding with CELP and the excitation signal of the low-frequency signal; performing audio decoding of the encoded audio frame based on the verification result; and performing frequency domain extension (FD) decoding using the result of performing audio decoding.

Performing decoding in an FD with extension may include performing energy inverse quantization using the same codebook at different bit rates.

Useful Results

According to an embodiment of the present invention, it is possible to effectively expand the high frequency range by extracting the main signal of the input signal and adjusting the energy of the input signal using the key of the high frequency range of the input signal and the key of the main signal.

Brief Description of the Drawings

FIG. 1 is a block diagram of an encoding device and a decoding device in accordance with an embodiment of the present invention.

FIG. 2A is a block diagram of an entire structure of an encoder in accordance with an embodiment of the present invention.

FIG. 2B is a block diagram of an entire structure of an encoder in accordance with another embodiment of the present invention.

FIG. 2C is a block diagram of a frequency domain encoder (FD) included in an encoding device according to an embodiment of the present invention.

FIG. 2D is a block diagram of the entire structure of an encoder in accordance with another embodiment of the present invention.

FIG. 3 is a block diagram of a basic encoder included in an encoding device according to an embodiment of the present invention.

FIG. 4 is a block diagram of an extension encoder included in an encoder in accordance with an embodiment of the present invention.

FIG. 5 is a block diagram of an extension encoder included in an encoder in accordance with another embodiment of the present invention.

FIG. 6 is a block diagram of a main signal generator included in an extension encoder according to an embodiment of the present invention.

FIG. 7 is a block diagram of a coefficient estimator included in an extension encoder in accordance with an embodiment of the present invention.

FIG. 8 is a flowchart illustrating an operation of an energy quantizer according to an embodiment of the present invention.

FIG. 9 is a diagram illustrating an energy quantization process in accordance with an embodiment of the present invention.

FIG. 10 is a diagram illustrating a process for generating an artificial signal in accordance with an embodiment of the present invention.

FIG. 11A and 11B respectively illustrate envelope estimation windows in accordance with embodiments of the present invention.

FIG. 12A is a block diagram of a decoding apparatus according to an embodiment of the present invention.

FIG. 12B is a block diagram of a decoding apparatus in accordance with another embodiment of the present invention.

FIG. 12C is a block diagram of an FD decoder included in a decoding apparatus according to an embodiment of the present invention.

FIG. 12D is a block diagram of a decoding apparatus in accordance with another embodiment of the present invention.

FIG. 13 is a block diagram of an extension decoder included in a decoding apparatus according to an embodiment of the present invention.

FIG. 14 is a flowchart illustrating the operation of an inverse quantizer included in an extension decoder in accordance with an embodiment of the present invention.

FIG. 15A is a flowchart illustrating an encoding method in accordance with an embodiment of the present invention.

FIG. 15B is a flowchart illustrating an encoding method in accordance with another embodiment of the present invention.

FIG. 15C is a flowchart illustrating an encoding method in accordance with another embodiment of the present invention.

FIG. 16A is a flowchart illustrating a decoding method in accordance with an embodiment of the present invention.

FIG. 16B is a flowchart illustrating a decoding method in accordance with another embodiment of the present invention.

FIG. 16C is a flowchart illustrating a decoding method in accordance with another embodiment of the present invention.

FIG. 17 is a structural diagram of an entire structure of an encoding device in accordance with another embodiment of the present invention.

FIG. 18 is a flowchart illustrating the operation of an energy quantizer included in an encoder in accordance with another embodiment of the present invention.

FIG. 19 is a diagram illustrating an energy quantization process using an unequal discharge distribution method according to an embodiment of the present invention.

FIG. 20 is a diagram illustrating vector quantization using intra-frame prediction in accordance with an embodiment of the present invention.

FIG. 21 is a diagram illustrating an energy quantization process using a frequency weighting method in accordance with another embodiment of the present invention.

FIG. 22 is a diagram illustrating vector quantization using multi-stage separate vector quantization and intra-frame prediction in accordance with an embodiment of the present invention.

FIG. 23 is a diagram illustrating the operation of an inverse quantizer included in a decoding apparatus in accordance with another embodiment of the present invention.

FIG. 24 is a block diagram of an entire structure of an encoding apparatus in accordance with another embodiment of the present invention.

FIG. 25 is a diagram illustrating binary signal streams in accordance with an embodiment of the present invention.

FIG. 26 is a diagram illustrating a method of performing frequency allocation for each frequency band in accordance with an embodiment of the present invention.

FIG. 27 is a diagram illustrating frequency bands used in an encoder in an FD or a decoder in an FD in accordance with an embodiment of the present invention.

Embodiments of the invention

Below will be described in detail typical embodiments of the present invention with reference to the accompanying drawings.

FIG. 1 is a block diagram of an encoder 101 and a decoder 102 in accordance with an embodiment of the present invention.

Encoding device 101 may generate a main signal (or basic signal) of the input signal and transmit the main signal to decoding device 102. The main signal is generated based on the low-frequency signal of the input signal. The main signal may be an excitation signal to expand the high frequency range, since the main signal is obtained by whitening the envelope information of the low-frequency signal. The decoding device 102 may restore the input signal from the main signal. In other words, the encoding device 101 and the decoding device 102 perform the extension of the ultra-wide frequency band (SWB BWE). In more detail, using the SWB BWE, a signal corresponding to the high frequency range from 6.4 to 16 kHz corresponding to the SWB can be generated based on a decoded wideband (WB) signal corresponding to the low frequency range from 0 to 6.4 kHz. Here 16 kHz may vary according to circumstances. The decoded WB signal may be generated using a speech codec in accordance with linear code-excited prediction (CELP) based on a linear prediction domain (LPD) or by performing quantization in the frequency domain. An example of a method for performing quantization in the frequency domain may include advanced audio coding (AAC) based on a modified discrete cosine transform (MDCT).

The operations of the encoder 101 and the decoder 102 are described in detail below.

FIG. 2A is a block diagram of an entire structure of an encoder 101 according to an embodiment of the present invention.

Referring to FIG. 2A, the encoder 101 may include a downsampler 201, a base encoder 202, a frequency converter 203, and an extension encoder 204.

For broadband (WB) coding, the downsampler 201 may lower the sampling frequency of the input signal. Typically, an input signal, such as an ultra-wideband (SWB) signal, has a sampling frequency of 32 kHz and is converted to a signal having a sampling frequency suitable for WB coding. For example, a downsampler 201 may downsample an input signal having a sampling frequency of, for example, 32 kHz, to a signal having a sampling frequency of, for example, 12.8 kHz.

Base encoder 202 may perform basic coding of the downsampled input signal. In other words, the base encoder 202 can perform WB encoding. For example, the base encoder 202 may perform WB encoding based on the CELP method.

The frequency converter 203 may perform frequency conversion on the input signal. For example, the frequency converter 203 may perform a fast Fourier transform (FFT) or MDCT to perform frequency conversion on an input signal. It is further assumed that MDCT is used.

Encoder 204 with extension can perform extension coding using the main input signal in the frequency domain. That is, encoder 204 with extension can perform BWE SWB encoding based on an input signal in the frequency domain. In this case, the encoder 204 with the extension does not receive encoding information, which will be described below with reference to FIG. four.

Also, the encoder 204 with the extension can perform wideband coding based on the characteristics of the input signal and based on the main signal of the input signal in the frequency domain. In this case, the encoder 204 with the extension may be implemented, as illustrated in FIG. 4 or 5, in accordance with the source of the characteristics of the input signal.

The operation of the encoder 204 with the extension will be described in detail below with reference to FIG. 4 and FIG. 5.

Thus, the upper path and lower path in FIG. 2A indicate a base coding process and a bandwidth extension coding process, respectively. Information about the energy of the input signal may be transmitted to the decoding device 102 by encoding the SWB BWE.

FIG. 2B is a block diagram of an entire structure of an encoder 101 in accordance with another embodiment of the present invention.

Referring to FIG. 2B, encoder 101 may include a signal classification unit 205, a CELP encoder 206, an extension in time domain (TD) 207, a frequency converter 208, and a frequency domain encoder 209 (FD).

Signal classification section 205 determines the encoding mode of the input signal based on the characteristics of the input signal. In the current embodiment, the encoding mode may be an encoding method.

For example, the signal classification block 205 may determine the coding mode of the input signal based on the characteristics of the time domain and the characteristics of the frequency domain of the input signal. Also, the signal classification block 205 determines that it is necessary to perform CELP encoding of the input signal when the characteristics of the input signal is a speech signal, and determines that it is necessary to FD-encode the input signal when the characteristics of the input signal is an audio signal.

However, the input to the signal classification block 205 may be a signal that has a reduced sampling rate by a downsampler (not shown). For example, in accordance with the current embodiment, the input signal may be a signal having a sampling frequency of 12.8 kHz or 16 kHz as a result of resampling a signal having a sampling frequency of 32 kHz or 48 kHz. Resampling may be downsampling.

As described above with reference to FIG. 2A, a signal having a sampling frequency of 32 kHz may be an ultra-wideband (SWB) signal. The SWB signal may be a full-range (FB) signal. A signal having a sampling frequency of 16 kHz may be a WB signal.

Signal classification section 205 may determine the coding mode of the low frequency signal corresponding to the low frequency range of the input signal as CELP mode or FD mode based on the characteristics of the low frequency signal.

If the encoding mode of the input signal is defined as the CELP mode, then the CELP encoder 206 performs CELP encoding of the low-frequency signal of the input signal. For example, a CELP encoder 206 can extract an excitation signal from a low frequency input signal and quantize the extracted excitation signal based on the contribution of the constant codebook and the contribution of the adaptive codebook corresponding to the pitch information.

However, the present invention is not limited to this, and the CELP encoder 206 can further extract linear prediction coefficients (LPCs) from the low frequency signal of the input signal, quantize the extracted LPC, and extract the drive signal using the quantized LPC.

Also, according to the current embodiment, the CELP encoder 206 may perform CELP encoding of the low frequency signal of the input signal in accordance with a different encoding mode in accordance with the characteristics of the low frequency signal of the input signal. For example, a CELP encoder 206 may perform CELP encoding of a low-frequency input signal in accordance with one of, for example, a tone encoding mode, a non-tone encoding mode, a transition encoding mode, and a general encoding mode.

When a low-frequency signal of an input signal is subjected to coding with CELP, the TD encoder 207 with extension performs coding with the extension of the high-frequency signal of the input signal. For example, a TD encoder 207 with extension quantizes the LPC of a high frequency signal corresponding to the high frequency range of the input signal. In this case, the encoder 207 in the TD with the extension can extract the LPC of the high frequency input signal and quantize the extracted LPC. Otherwise, the TD encoder 207 with the extension may generate an LPC of the high frequency input signal using the excitation signal of the low frequency input signal.

Thus, the encoder 207 in the TD extension can be a high frequency encoder in the TD extension, but the present invention is not limited to it.

If the encoding mode of the input signal is defined as the FD encoding mode, then the frequency converter 208 performs frequency conversion on the input signal. For example, the frequency converter 208 may perform a frequency conversion that includes overlapping frames, such as MDCT, on an input signal, but the present invention is not limited to this

Encoder 209 in FD performs FD encoding of the converted frequency input signal. For example, an encoder 209 in an FD may perform FD encoding of a frequency spectrum converted by a frequency converter 208. The encoder 209 in the FD will be described in detail below with reference to FIG. 2C.

According to the current embodiment, the encoder 101 may output a stream of binary signals by encoding an input signal, as described above. For example, a binary stream may include a header and a payload.

The header may include encoding mode information indicating an encoding mode used to encode an input signal. The payload may include information in accordance with the encoding mode used to encode the input signal. If the input signal is encoded in accordance with the CELP mode, then the payload may include CELP information and high frequency extension TD information. If the input signal is encoded in accordance with the FD mode, then the payload may include prediction data and FD information.

However, in the binary signal stream in accordance with the current embodiment, the header may further include previous frame mode information to correct for a frame error that may occur. For example, if the encoding mode of the input signal is defined as the FD mode, then the header may further include information about the mode of the previous frame, which will be described in detail below with reference to FIG. 25.

Thus, the encoder 101 in accordance with the current embodiment switches to using the CELP mode or the FD mode in accordance with the characteristics of the input signal, thereby encoding the input signal in a suitable manner in accordance with the characteristics of the input signal. Also, the encoder 101 uses the FD mode in accordance with the definition of the signal classification block 205, thereby performing encoding in an appropriate manner in a high speed environment.

FIG. 2C is a block diagram of an FD encoder 209 in accordance with an embodiment of the present invention.

Referring to FIG. 2C, the FD encoder 209 may include a normalization encoder 2091, a pulse factor encoder 2092, an additional noise information generator 2093, and an extension FD encoder 2094.

Normalization encoder 2091 extracts energy from each frequency band of the input signal converted by the frequency converter 208, and quantizes the extracted energy. Also, normalization encoder 2091 may perform scaling based on the extracted energy. In this case, a scaled energy value can be quantized. For example, the energy value in accordance with the current embodiment can be obtained using a measurement method for measuring energy or power having a proportional relationship with the energy of the frequency band.

The normalized information, that is, the result of the quantization performed by the normalization encoder 2091, can be included in the binary signal stream and transmitted along with the binary signal stream to decoding device 102.

For example, normalization encoder 2091 divides the frequency spectrum corresponding to the input signal into a predetermined number of frequency bands, extracts energy from the frequency spectrum for each frequency band, and quantizes the extracted energies. The quantized value can be used to normalize the frequency spectrum.

Normalization encoder 2091 may further encode a quantized value.

The factorial pulse encoder 2092 can perform factorial pulse encoding (FPC) of the value obtained by scaling the converted input signal using the quantized normalization value. In other words, the factorial pulse encoder 2092 can perform FPC values of the spectrum normalized by the normalization encoder 2091.

For example, factorial-pulse encoder 2092 assigns a certain number of available bits to each frequency band and performs FPC normalized spectrum values in accordance with the assigned number of bits. In this case, the number of bits assigned to each frequency band may be determined in accordance with the target bit rate. Also, the factorial-pulse encoder 2092 can calculate the number of bits to be assigned to each frequency band using the normalization coding value quantized by the normalization encoder 2091. In this case, the factorial-pulse encoder 2092 can perform the FPC spectrum with a converted frequency, rather than a normalized spectrum.

An additional noise information generator 2093 generates additional noise information in accordance with the implementation of the FPC. For example, the additional noise information generator 2093 generates a suitable noise level based on the result of performing the FPC frequency spectrum using a factorial-pulse encoder 2092.

In this case, the additional noise information generated by the additional noise information generator 2093 may be included in the binary signal stream so that the decoding side can access the additional noise information to perform noise filling.

Encoder 2094 in FD with extension performs extension coding of the high-frequency signal of the input signal. More specifically, the FD Encoder 2094 with extension performs high-frequency expansion using the low-frequency spectrum.

For example, an FD encoder 2094 with extension quantizes frequency domain energy information from a high frequency signal corresponding to the high frequency range of the input signal. In this case, the FD encoder 2094 with the extension can split the frequency spectrum corresponding to the input signal into a predetermined number of frequency bands, obtain an energy value from the frequency spectrum for each frequency band, and perform multi-stage vector quantization (MSVQ) using the energy value. MSVQ may be multi-stage vector quantization.

More specifically, an FD encoder 2094 with extension can perform vector quantization (VQ) by collecting information about the energy of odd frequency bands among a predetermined number of frequency bands, obtain a predicted error in an even frequency band based on a quantized value in accordance with the result of vector quantization, and perform vector quantization of the obtained predicted error in the next step.

However, the present invention is not limited to this, and an FD encoder 2094 with extension can perform vector quantization by collecting information about the energy of even frequency bands among a predetermined number of frequency bands and obtain the predicted error in the odd frequency band using the quantized value in accordance with the result of vector quantization .

That is, the encoder 2094 in the FD with the extension receives the predicted error in the (n + 1) ^th frequency band from the quantized value obtained by performing vector quantization of the n- ^th frequency band and the quantized value obtained by performing vector quantization of the (n + 2) ^-th band frequencies. Here, "n" is a natural number.

Also, in order to perform vector quantization by collecting energy information, an FD encoder 2094 with extension can simulate a method of generating an excitation signal in a predetermined frequency band and can adjust energy when the characteristics of the excitation signal in accordance with the simulation result differ from the characteristics of the original signal in advance set frequency band. In this case, the characteristics of the excitation signal in accordance with the simulation result and the characteristics of the original signal may include at least one of tonality and noise figure, but the present invention is not limited to this. Thus, an increase in noise can be prevented when the decoding side decodes the actual energy.

Encoder 2094 in the FD with the extension can use multi-mode bandwidth extension, which uses various methods of generating the excitation signal in accordance with the characteristics of the high-frequency signal of the input signal. For example, an FD encoder 2094 with an extension according to the characteristics of the input signal may use one of the normal mode, harmonics mode, and noise mode for each frame to generate an excitation signal.

Also, in accordance with the current embodiment, the encoder 2094 in the FD with the extension can generate a signal of a frequency band that changes in accordance with the bit rate. That is, the high-frequency range corresponding to the high-frequency signal that the encoder 2094 in the FD extension encodes with the extension can be set differently in accordance with the bit rate.

For example, an FD encoder 2094 with extension can be used to generate a signal corresponding to a frequency band of approximately 6.4 to 14.4 kHz, at a bit rate of 16 kbps and to generate a signal corresponding to a frequency band of approximately 8 to 16 kHz, at a bit rate that is greater than or equal to 16 kbit / s. Also, an encoder 2094 in FD with extension performs extension coding of a high-frequency signal corresponding to a frequency band of approximately 6.4 to 14.4 kHz at a bit rate of 16 kbps and performs extension coding of a high-frequency signal corresponding to a frequency band of approximately 8 to 16 kHz, at a bit rate that is greater than or equal to 16 kbit / s.

In this case, according to the current embodiment, the FD encoder 2094 with extension can perform energy quantization by sharing the same codebook at different bit rates, which will be described in detail below with reference to FIG. 26.

If a constant frame is entered into the encoder 209 in the FD, then the normalization encoder 2091, the pulse factor encoder 2092, the additional noise information generator 2093, and the extension encoder 2094 in the FD can operate in the encoder 209 in the FD.

However, when a transition frame is introduced, the encoder 2094 in the FD extension may not work. In this case, the normalization encoder 2091 and the factorial pulse encoder 2092 can set a higher Fcore value of the upper band at the frequency band over which FPC is to be performed than when entering a constant frame. The Fcore value of the upper band will be described in detail below with reference to FIG. 27.

FIG. 2D is a block diagram of the entire structure of an encoder 101 in accordance with another embodiment of the present invention.

Referring to FIG. The 2D encoder 101 may include a signal classification block 210, an LPC encoder 211, a CELP encoder 212, a TD encoder 213 with extension, an audio encoder 214, and an FD encoder 215 with an extension.

Block 210 classification of the signal determines the encoding mode of the input signal in accordance with the characteristics of the input signal. According to the current embodiment, the encoding mode may be an encoding method.

For example, the signal classification block 210 determines an encoding mode of an input signal based on characteristics of a time domain and characteristics of a frequency domain of an input signal. Also, the signal classification unit 205 may determine that it is necessary to perform CELP encoding of the input signal when the characteristics of the input signal is a speech signal, and determine that it is necessary to perform audio coding of the input signal when the characteristics of the input signal is an audio signal.

The LPC encoder 211 extracts the LPC from the low frequency signal of the input signal and quantizes the LPC. For example, in accordance with the current embodiment, the LPC encoder 211 may use, for example, grid geometry quantization (TCQ), MSVQ, or trellis vector quantization (LVQ) to quantize LPCs, but the present invention is not limited to them.

More specifically, the LPC encoder 211 may resample, for example, an input signal having a sampling frequency of 32 kHz or 48 kHz to extract an LPC from a low frequency signal of an input signal having a sampling frequency of 12.8 kHz or 16 kHz.

As described above with reference to FIG. 2A and 2B, a signal having a sampling frequency of 32 kHz may be a SWB signal. The SWB signal may be an FB signal. Also, a signal having a sampling frequency of 16 kHz may be a WB signal.

The LPC encoder 211 may further extract an LPC drive signal using a quantized LPC, but the present invention is not limited to this.

If the encoding mode of the input signal is defined as CELP mode, then the CELP encoder 212 encodes with CELP the LPC excitation signal extracted using the LPC. For example, a CELP encoder 212 may quantize the LPC drive signal based on the contribution of the constant codebook and the contribution of the adaptive codebook corresponding to the fundamental frequency information. In this case, the LPC drive signal may be generated by at least one of the CELP encoder 212 and the LPC encoder 211.

According to the current embodiment, the CELP encoder 212 may also perform CELP encoding in accordance with various encoding modes in accordance with the characteristics of the low-frequency signal of the input signal. For example, a CELP encoder 206 may perform CELP encoding of a low-frequency input signal using one of the encoding tone mode, non-tone encoding mode, transient encoding mode, or general encoding mode.

Encoder 213 in TD with extension performs extension coding of the high-frequency input signal when the LPC drive signal of the low-frequency input signal is encoded with CELP.

For example, encoder 213 in TD with the extension quantizes the LPC of the high-frequency signal of the input signal. In this case, the encoder 213 in the TD with the extension can extract the LPC of the high-frequency input signal using the LPC drive signal of the low-frequency input signal.

Thus, the encoder 213 in the TD with the extension may be a high frequency encoder in the TD with the extension, but the present invention is not limited to them.

If the encoding mode of the input signal is defined as the audio encoding mode, then the audio encoder 214 audio encodes the LPC drive signal extracted using the LPC.

For example, audio encoder 214 may perform frequency conversion on the LPC drive signal and quantize the converted LPC drive signal.

Here, when the audio encoder 214 performs frequency conversion, the audio encoder 214 may use a frequency conversion method that does not include overlapping frames, such as discrete cosine transform (DCT). Also, audio encoder 214 may quantize the spectrum of the converted frequency excitation signal in accordance with FPC or trellis VQ (LVQ).

In addition, if the audio encoder 214 has spare bits to perform quantization of the LPC drive signal, then the audio encoder 214 may further quantize based on the encoding information in TD in the contribution of the constant codebook and the contribution of the adaptive codebook.

When the LPC drive signal of the low frequency signal of the input signal is subjected to audio coding, the encoder 215 in the FD with extension performs coding with the extension of the high frequency signal of the input signal. In other words, the encoder 215 in the FD with the extension can perform high-frequency expansion using the low-frequency spectrum.

For example, an FD encoder 215 with extension quantizes frequency domain energy information of a high frequency signal corresponding to the high frequency range of the input signal. In this case, the encoder 215 in the FD with the extension can generate the frequency spectrum using a frequency conversion method, for example MDCT, divide the frequency spectrum into a predetermined number of frequency bands, obtain the energy of the frequency spectrum for each frequency band and perform MSVQ using energy. Here, the MSVQ may be multi-stage vector quantization.

More specifically, the FD encoder 215 with extension can perform vector quantization by collecting information about the energy of odd frequency bands among a predetermined number of frequency bands, obtain a predicted error in an even frequency band based on a quantized value in accordance with a vector quantization result, and perform vector quantization of the predicted errors in the next step.

However, the present invention is not limited to this, and the FD encoder 215 with extension can perform vector quantization by collecting information about the energy of even frequency bands among a predetermined number of frequency bands and obtain the predicted error in the odd frequency band using the quantized value in accordance with the result of vector quantization .

That is, the encoder 215 in the FD with the extension receives the predicted error in the (n + 1) ^th frequency band using the quantized value obtained by performing vector quantization of the n- ^th frequency band and the quantized value obtained by performing vector quantization of the (n + 2) ^th frequency bands. Here, "n" is a natural number.

Also, in order to perform vector quantization by collecting energy information, the FD encoder 215 with extension can simulate a method of generating an excitation signal in a predetermined frequency band and can control energy when the characteristics of the excitation signal in accordance with the simulation result differ from the characteristics of the original signal in advance set frequency band. In this case, the characteristics of the excitation signal in accordance with the simulation result and the characteristics of the original signal may include at least one of tonality and noise figure, but the present invention is not limited to this. Thus, an increase in noise can be prevented when the decoding side decodes the actual energy.

Encoder FD 215 may use multi-mode bandwidth extension, which uses various methods of generating an excitation signal in accordance with the characteristics of the high-frequency signal of the input signal. For example, an FD encoder 215 with an extension according to the characteristics of the input signal may generate an excitation signal using one of the normal mode, harmonic mode, transition mode, or noise mode for each frame. In transient mode, temporal envelope information may also be quantized.

Also, in accordance with the current embodiment, the encoder 215 in the FD with the extension can generate a signal of a frequency band that changes in accordance with the bit rate. In other words, the high frequency range corresponding to the high frequency signal that the encoder 215 in the FD with extension encodes with the extension can be set differently according to the bit rate.

For example, encoder 215 in an FD with extension can be used to generate a signal corresponding to a frequency band of approximately 6.4 to 14.4 kHz, at a bit rate of 16 kbps and to generate a signal corresponding to a frequency band of approximately 8 to 16 kHz, at a bit rate that is greater than or equal to 16 kbit / s. Also, the encoder 215 in the FD with extension can perform extension coding of a high frequency signal corresponding to a frequency band of approximately 6.4 to 14.4 kHz at a bit rate of 16 kbps and perform encoding with extension of a high frequency signal corresponding to a frequency band of approximately from 8 to 16 kHz, at a bit rate that is greater than or equal to 16 kbit / s.

In this case, according to the current embodiment, the FD encoder 215 with extension can perform energy quantization by sharing the same codebook at different bit rates, which will be described in detail below with reference to FIG. 26.

In the current embodiment, the encoder 101 may encode the input signal as described above and output it as an encoded binary stream. For example, a binary stream includes a header and a payload.

In this case, the header may include encoding mode information indicating an encoding mode used to encode the input signal. The payload may include CELP information and high frequency extension TD information when the input signal is encoded using the CELP mode, and may include prediction data, audio encoding information and high frequency extension information FD when the input signal is encoded using the audio encoding mode.

Encoder 101 may switch to use the CELP mode or audio coding mode in accordance with the characteristics of the input signal. Thus, a suitable encoding mode may be performed in accordance with the characteristics of the input signal. In addition, the encoder 101 may use the FD mode in accordance with the definition of the signal classification block 210, thereby performing encoding in an appropriate manner in a low speed environment.

FIG. 3 is a block diagram of a basic encoder 202 in an encoder 101 in accordance with an embodiment of the present invention.

Referring to FIG. 3, the base encoder 202 may include a signal classification unit 301 and an encoder 302.

Signal classification section 301 may classify the characteristics of the downsampled input signal, for example 12.8 kHz. In other words, the signal classification section 301 may classify the encoding modes of the input signal as different encoding modes in accordance with the characteristics of the input signal. For example, in accordance with the ITU-T G.718 codec, the signal classification unit 301 can classify speech coding modes as a coding mode with vocals, a coding mode without vocals, a transient coding mode, and a general coding mode. The encoding mode without vocals is intended for encoding frames without vocals and most inactive frames.

Encoder 302 may perform coding optimized for the characteristics of the input signal, classified by block 301 signal classification.

FIG. 4 is a block diagram of an extension encoder 204 in an encoder 101 in accordance with an embodiment of the present invention.

Referring to FIG. 4, the extension encoder 204 may include a main signal generator 401, a coefficient estimator 402, an energy extraction module 403, an energy controller 404, and an energy quantizer 405. For example, encoder 204 with the extension may evaluate the energy control coefficient without receiving information about the encoding mode. As another example, an encoder 204 with an extension may estimate an energy control coefficient using an encoding mode. Information about the encoding mode may be received from the base encoder 202.

The main signal generator 401 can generate the main signal of the input signal using the frequency spectrum of the input signal in the frequency domain. The main signal indicates a signal for performing the SWB BWE based on the WB signal. In other words, the main signal indicates a signal that makes up the exact structure of the low frequency range. The process of generating the main signal will be described in detail below with reference to FIG. 6.

For example, coefficient estimation module 402 may estimate the energy control coefficient using the main signal. That is, the encoding device 101 transmits the energy information of the input signal to generate the signal of the SWB region in the decoding device 102. In this case, the coefficient estimator 402 can estimate the energy control coefficient, which is a parameter for regulating the energy to control energy information from a perceptual . The process of estimating the energy control coefficient will be described in detail below with reference to FIG. 7.

As another example, coefficient estimation module 402 may estimate the energy control coefficient using the characteristics of the main signal and the input signal. In this case, the characteristics of the input signal may be received from the base encoder 202.

An energy extraction module 403 can extract energy from an input signal in a frequency band. The extracted energy is transmitted to the decoding device 102. Energy can be extracted in each frequency band.

An energy controller 404 may adjust the energy extracted from the input signal using an energy control coefficient. In other words, the energy controller 404 can adjust the energy by applying an energy control coefficient to the energy extracted in each frequency band.

An energy quantizer 405 can quantize an adjustable energy. Energy can be converted to decibels, and then quantized. In particular, energy quantizer 405 can calculate global energy, which is total energy, and scalarly quantize global energy and the differences between global energy and the energy extracted in each frequency band. Otherwise, the energy extracted from the first frequency band is directly quantized, and then the difference between the energy extracted in each of the frequency bands besides the first frequency band and the energy extracted in the previous frequency band can be quantized. Otherwise, the energy quantizer 405 can directly quantize the energy extracted in each frequency band without using the differences between the energies extracted in the frequency bands. Scalar or vector quantization can be used when the extracted energy in each frequency band is directly quantized. An energy quantizer 405 will be described in detail below with reference to FIG. 8 and 9.

FIG. 5 is a block diagram of an extension encoder 204 in an encoder 101 in accordance with another embodiment of the present invention.

Referring to FIG. 5, an encoder 204 with an extension compared to an encoder 204 with an extension from FIG. 4 may further include a signal classification unit 501. For example, coefficient estimation module 402 may estimate the energy control coefficient using the characteristics of the main signal and the input signal. In this case, the characteristics of the input signal may be received from the signal classification unit 501, and not from the base encoder 202.

Block 501 signal classification can classify the input signal, for example, 32 kHz and the spectrum of MDCT, in accordance with the characteristics of the input signal. In more detail, the signal classification unit 501 can classify the encoding modes of the input signal as different encoding modes based on the characteristics of the input signals.

By classifying the input signal according to the characteristics of the input signal, the energy control coefficient can only be estimated from signals suitable for performing the process of evaluating the energy control coefficient, and can control the energy. For example, it may not be appropriate to carry out the process of evaluating the energy control coefficient over a signal that does not contain a tone component, for example, a noise signal or a signal without vocals. In this case, if the encoding mode of the input signal is classified as encoding mode without vocals, then the extension encoder 204 can perform extension band encoding without performing an estimation of the energy control coefficient.

The main signal generator 401, coefficient estimation module 402, energy extraction module 403, energy controller 404, and energy quantizer 405, illustrated in FIG. 5 are as described above with reference to FIG. four.

FIG. 6 is a block diagram of a main signal generator 401 included in an extension encoder 204 in accordance with an embodiment of the present invention.

Referring to FIG. 6, the main signal generator 401 may include an artificial signal generator 601, an envelope estimation module 602, and an envelope application unit 603.

The artificial signal generator 601 can generate an artificial signal corresponding to the high frequency range by copying and folding the low frequency range of the input signal in the frequency band. In other words, the artificial signal generator 601 can generate an artificial signal in the SWB region by copying the low-frequency spectrum of the input signal in the frequency domain. The artificial signal generation process will be described in detail below with reference to FIG. 6.

Envelope estimator 602 may estimate the envelope of the main signal using a window. The envelope of the main signal can be used to exclude envelope information about the low frequency range included in the frequency spectrum of the artificial signal in the SWB region. An envelope with a specific frequency index can be determined using frequency spectra before and after a specific frequency. The envelope of the main signal can also be estimated by means of a moving average. For example, if MDCT is used for frequency conversion, then the envelope of the main signal can be estimated by the absolute value of the frequency spectrum, which is converted by MDCT.

In this case, the envelope estimator 602 may create whitening strips, calculate an average frequency value in each of the whitening strips, and estimate the average frequency value of the whitening stripe as an envelope of frequencies belonging to the whitening stripe. The number of frequency spectra belonging to the whitening band can be set less than the number of bands from which the energy is extracted.

If the average frequency value calculated in each of the whitening bands is estimated as an envelope of the frequency belonging to the whitening band, then the envelope estimator 602 may transmit information indicating whether the number of frequency spectra belonging to the whitening bands is large or small to control the degree of unevenness main signal. For example, envelope estimation module 602 may transmit such information depending on two types of the fact that the number of frequency spectra is eight or three. If the number of frequency spectra is three, then the degree of unevenness of the main signal can be higher than when the number of frequency spectra is eight.

Otherwise, the envelope estimator 602 may not transmit information indicating whether the number of frequency spectra belonging to the whitening bands is large or small, and may determine the degree of unevenness of the main signal in accordance with the encoding mode used by the base encoder 202. The base encoder 202 may classify input signal encoding mode as encoding mode with vocals, encoding mode without vocals, transient encoding mode or general encoding mode based on character IR input signal and may encode the input signal.

In this case, the envelope estimator 602 may control the number of frequency spectra belonging to the whitening bands based on the encoding mode in accordance with the characteristics of the input signal. For example, if an input signal is encoded according to a coding mode with vocals, then envelope estimator 602 can estimate the envelope of the main signal by creating three frequency spectra in the whitening band. If the input signal is encoded in accordance with a coding mode different from the coding mode with vocals, then the envelope estimator 602 can estimate the envelope of the main signal by creating three frequency spectra in the whitening band.

Envelope application block 603 may apply the estimated envelope to the artificial signal. This process is consistent with the bleaching process. An artificial signal can be aligned using an envelope. Envelope application unit 603 can generate a main signal by dividing the artificial signal according to the envelope of each of the frequency indices.

FIG. 7 is a block diagram of a coefficient estimator 402 included in an extension encoder 204 in accordance with an embodiment of the present invention.

Referring to FIG. 7, the coefficient estimator 402 may include a first tonality calculator 701, a second tonality calculator 702, and a coefficient calculator 703.

The first tonality calculator 701 can calculate the tonality of the high frequency range of the input signal in the frequency domain. In other words, the first tonality calculator 701 can calculate the tonality of the SWB region, which is the high frequency range of the input signal in the frequency domain.

The second tonality calculator 702 may calculate the tonality of the main signal.

Keys can be calculated by measuring spectral unevenness. In more detail, tonality can be calculated using Equation (1) below. Spectral non-uniformity can be measured using the relationship between the geometric mean and arithmetic mean of the frequency spectrum.

The coefficient calculator 703 can calculate the energy control coefficient using the key of the high frequency range of the input signal and the key of the main signal. In this case, the energy regulation coefficient can be calculated using:

where "α" indicates the coefficient of energy regulation, "T _o " indicates the tonality of the input signal, and "T _b " indicates the tonality of the main signal. Also, “N _b ” denotes a noise figure that indicates the degree of content of the noise component in the signal.

The energy control coefficient can be calculated using:

$$\alpha =\frac{T_h}{T_o}$$

$$(3)$$

$$$$

A coefficient calculator 703 can calculate an energy control coefficient for each frequency band. The calculated energy control coefficient can be applied to the energy of the input signal. In this case, the energy control coefficient can be applied to the energy of the input signal when the energy control coefficient is less than a predetermined threshold energy control coefficient.

FIG. 8 is a flowchart illustrating an operation of an energy quantizer 405 in accordance with an embodiment of the present invention.

In step S801, the energy quantizer 405 can pre-process the energy vectors using an energy control coefficient and select a sub-vector of the pre-processed energy vector. For example, energy quantizer 405 can subtract the average of energy vectors from each of the energy vectors or calculate the weight relative to the importance of each of the energy vectors. In this case, the weight can be calculated so that the quality of the synthesized sound can increase as much as possible.

Also, energy quantizer 405 may suitably select a subvector of the energy vector based on coding efficiency. Also, energy quantizer 405 may select a subvector in the same time interval to increase interpolation efficiency.

For example, energy quantizer 405 may select a subvector in accordance with equation (4) below.

k × n (n = 0, ..., N), k≥2, N denotes the largest integer that is less than the dimension of the vector (4)

In this case, if k = 2, then only even numbers are selected.

In step S802, the energy quantizer 405 quantizes and quantizes the selected subvector inversely. An energy quantizer 405 can quantize a subvector by selecting a quantization index to minimize the mean square error (MSE) calculated using equation (5) below.

An energy quantizer 405 may quantize a subvector using scalar quantization, vector quantization, TCQ, or LVQ. In vector quantization, an MSVQ or a separate VQ, or a separate VQ and a multi-stage VQ can be performed simultaneously. The quantization index is transmitted to decoding device 102.

When weights are calculated during pre-processing, energy quantizer 405 can calculate an optimized quantization index using weighted MSE (WMSE). In this case, the WMSE can be calculated using:

In step S803, the energy quantizer 405 may interpolate the remaining subvectors that are not selected.

In step S804, the energy quantizer 405 can calculate the interpolation errors, which are the differences between the interpolated remaining subvectors and the original subvectors that match the energy vectors.

In step S805, the energy quantizer 405 quantizes and inversely quantizes the interpolation error. In this case, the energy quantizer 405 may quantize the interpolation error using a quantization index to minimize MSE. An energy quantizer 405 may quantize an interpolation error using scalar quantization, vector quantization, TCQ, or LVQ. In this case, when vector quantization can be performed MSVQ or separate VQ, or separate VQ and MSVQ can be performed simultaneously. If weights are calculated during pre-processing, then energy quantizer 405 can calculate an optimized quantization index using WMSE.

In step S806, the energy quantizer 405 can calculate the remaining subvectors that are not selected by interpolating the quantized subvectors that are selected and calculate the value of the quantized energy by adding the quantized interpolation errors calculated in step S805. And the energy quantizer 405 can calculate the final quantized energy by repeatedly adding the average that is subtracted in the pre-processing during the pre-processing.

In MSVQ, the energy quantizer 405 performs quantization using K candidate subvectors to improve quantization performance based on the same codebook. If "K" is greater than or equal to "2", then energy quantizer 405 can determine the optimal candidate candidate subvectors by performing a distortion measurement. In this case, the measurement of distortion can be determined in accordance with one of the following two methods.

In the first, the energy quantizer 405 may generate an index set to minimize MSE or WMSE for each of the candidate sub-vectors at each of the stages, and select a candidate subvector having the smallest sum of MSE or WMSE at all stages of the candidate sub-vectors. In this case, the amount of computation is small.

In the second, the energy quantizer 405 can generate an index set to minimize MSE or WMSE for each of the candidate sub-vectors at each step, restore the energy vector by inverse quantization, and select a candidate sub-vector to minimize MSE or WMSE between the reconstructed energy vector and the original vector energy. In this case, the amount of computation increases due to recovery of the energy vector, but the performance is better because MSEs are calculated using actually quantized values.

FIG. 9 is a diagram illustrating an energy quantization process in accordance with an embodiment of the present invention.

Referring to FIG. 9, the energy vector represents 14 dimensions. In a first step, the energy quantizer 405 selects subvectors corresponding to dimension 7 by selecting even subvectors of the energy vector. In a first step, the energy quantizer 405 uses two-vector quantization of the second step to improve performance.

An energy quantizer 405 quantizes in a second step using an error signal of a first step. The energy quantizer 405 calculates the interpolation error by inverse quantization of the selected subvectors and quantizes the interpolation error by vector quantization of the third stage, divided into two.

FIG. 10 is a diagram illustrating a process for generating an artificial signal in accordance with an embodiment of the present invention.

Referring to FIG. 10, an artificial signal generator 601 can copy a frequency spectrum 1001 corresponding to a low frequency range from f _L to 6.4 kHz over the entire frequency band. The copied frequency spectrum 1001 is shifted into the frequency band from 6.4 to 12.8-f _L kHz. The frequency spectrum corresponding to the frequency band from 12.8-f _L to 16 kHz can be formed by folding the frequency spectrum corresponding to the frequency band from 6.4 to 12.8-f _L kHz. In other words, an artificial signal corresponding to the SWB region, which is a high frequency range, is generated from 6.4 to 16 kHz.

If MDCT is performed to form the frequency spectrum, then there is a correlation between f _L and 6.4 kHz. In more detail, when the frequency index MDCT corresponding to 6.4 kHz is an even number, the frequency index y f _{L is} also an even number. In contrast, if the MDCT frequency index corresponding to 4 kHz is an odd number, then the frequency index of f _{L is} also an odd number.

For example, when MDCT is applied for extracting the frequency spectrum 640 of the original input signal, the index corresponding to 6.4 kHz 256 is ^{the second} index (i.e., 6400/16000 * 640), i.e. an even number. In this case, f _{L is} also selected as an even number. In other words, 2 (50 Hz) or 4 (100 Hz) can be used for f _L. This process can also be used during the decoding process.

FIG. 11A and 11B respectively illustrate envelope estimation windows 1101 and 1102 in accordance with embodiments of the present invention.

With reference to FIG. 11A and 11B, the peak point on each of the windows 1101 and 1102 denotes a frequency index for estimating the current envelope. The current envelope of the main signal can be estimated using:

With reference to FIG. 11A and 11B, windows 1101 and 1102 can be used invariably, in which case no additional bits need to be transmitted. If window 1101 or 1102 is selectively used, information indicating whether window 1101 or 1102 was used to estimate the envelope needs to be expressed using bits and additionally transmitted to decoding device 102. Bits may be transmitted for each frequency band or may be transmitted immediately in one frame.

The weight is additionally added to the frequency spectrum corresponding to the current frequency index in order to estimate the envelope when using window 1102 compared to when using window 1101. Thus, the main signal generated using window 1102 is smoother than that generated using window 1101. The type of window from windows 1101 and 1102 can be selected by comparing each of the main signals generated by window 1101 and window 1102 with the frequency spectrum of the input signal. Otherwise, a window having a tonality that is closer to the tonality of the high frequency range can be selected from windows 1101 and 1102 by comparing the tonality of the high frequency range. Otherwise, a window having a large correlation with the high frequency range may be selected from windows 1101 and 1102 by comparing the correlation.

FIG. 12A is a block diagram of a decoding device 102 in accordance with an embodiment of the present invention.

The decoding process performed by the decoding device 102 of FIG. 12A is the inverse process to the process executed by the encoder 101 of FIG. 2A. Referring to FIG. 12A, the decoding apparatus 102 may include a base decoder 1201, an upsampler 1202, a frequency converter 1203, an extension decoder 1204, and a frequency inverter 1205.

The base decoder 1201 may perform basic decoding of the base-coding input signal contained in the binary stream. Through basic decoding, a signal having a sampling frequency of 12.8 kHz can be extracted.

An upsampler 1202 may increase the sampling rate of a basic decoded input signal. By upsampling, a signal having a sampling frequency of 32 kHz can be extracted.

Frequency converter 1204 may perform frequency conversion on an input signal with an increased sampling rate. In this case, the same frequency conversion as used in the encoder 101 can be used. For example, MDCT can be used.

Enhanced decoder 1204 can perform wideband decode using the input signal in the frequency band and the energy of the input signal contained in the binary signal stream. The operation of the decoder 1204 with the extension will be described in detail below with reference to FIG. 9.

The inverse frequency converter 1205 may perform the inverse frequency conversion on the result of performing bandwidth extension decoding. In other words, inverse frequency conversion may be the inverse of the frequency conversion performed by the frequency converter 1204. For example, the inverse frequency transform may be an inverse modified discrete cosine transform (IMDCT).

FIG. 12B is a block diagram of a decoding device 102 in accordance with another embodiment of the present invention.

The decoding process performed by the decoding device 102 of FIG. 12B is the inverse process to the process of FIG. 12A. Referring to FIG. 12B, the decoding apparatus 102 may include a mode information checking unit 1206, a CELP decoder 1207, an extension TD decoder 1208, an FD decoder 1209, and a frequency inverter 1210.

The mode information checking unit 1206 checks the mode information of each of the frames included in the binary signal stream. The binary signal stream may be a signal corresponding to the binary signal stream in accordance with the encoding result performed by the encoding device 101 transmitted to the decoding device 102.

For example, mode information checking unit 1206 analyzes mode information from a binary signal stream and performs an operation of switching to one of the decoding mode with CELP or FD decoding mode in accordance with the encoding mode of the current frame in accordance with the analysis result.

More specifically, the mode information checking unit 1206 can be switched with respect to each of the frames included in the binary signal stream, so that the frame encoded in accordance with the CELP mode can be decoded with CELP, and the frame encoded in accordance with the FD mode, can be decoded by FD.

A CELP decoder 1207 decodes a CELP frame encoded in accordance with the CELP mode based on the verification result. For example, a CELP decoder 1207 decodes an LPC included in a binary signal stream, decodes adaptive and constant codebook contributions, combines the decoding results, and generates a low-frequency signal corresponding to the decoded signal for the low frequency range.

Decoder 1208 in TD with the extension generates a decoded signal for the high frequency range using at least one of the result of decoding with CELP and the excitation signal of the low-frequency signal. In this case, the low-frequency excitation signal may be included in the binary signal stream. Also, the TD decoder 1208 with extension can use the LPC information about the high frequency signal included in the binary signal stream to generate a high frequency signal corresponding to the decoded signal for the high frequency range.

Also in accordance with the current embodiment, the TD decoder 1208 with extension can generate a decoded signal by combining a high frequency signal with a low frequency signal generated by a CELP decoder 1207. To generate the decoded signal, the decoder 1208 in TD with the extension can further convert the sampling frequencies of the low-frequency signal and the high-frequency signal to the same.

An FD decoder 1209 performs FD decoding of an FD encoded frame. Decoder 1209 in the FD may generate a frequency spectrum by decoding a stream of binary signals. Also, in accordance with the current embodiment, the decoder 1209 in the FD may perform decoding of the binary signal stream based on the mode information of the previous frame included in the binary signal stream. In other words, the FD decoder 1209 may perform FD decoding of FD-encoded frames based on the previous frame mode information included in the binary stream, as will be described in detail below with reference to FIG. 25. The decoder 1209 in the FD will be described in detail below with reference to FIG. 12C.

The inverse frequency converter 1210 performs inverse frequency conversion on the result of performing FD decoding. The inverse frequency converter 1210 generates a decoded signal by performing inverse frequency conversion on the frequency spectrum decoded in the FD. For example, inverse frequency converter 1210 may perform inverse MDCT, but the present invention is not limited thereto.

Accordingly, decoding device 102 may decode the binary signal stream based on the encoding modes of each of the frames in the binary signal stream.

FIG. 12C is a block diagram of an FD decoder 1209 included in decoding device 102 in accordance with an embodiment of the present invention.

The decoding process performed by the decoder 1209 in the FD of FIG. 12C is the inverse process to the process of FIG. 12B. Referring to FIG. 12C, the FD decoder 1209 may include a normalization decoder 12091, an FPC decoder 12092, a noise filling unit 12093, and an extension FD decoder 12094. Decoder 12094 in FD with extension may include a low-frequency decoder 12095 in FD with extension and high-frequency decoder 12096 in FD with extension.

The normalization decoder 12091 performs normalization decoding based on the normalization information about the binary stream. The normalization information may be information in accordance with the encoding result of the normalization encoder 2091 of FIG. 2C.

FPC decoder 12092 performs FPC decoding based on FPC binary stream information. The FPC information may be information in accordance with the encoding result of the factorial pulse encoder 209 of FIG. 2C.

For example, FPC decoder 12092 performs FPC decoding by assigning a certain number of bits available in each frequency band, similar to encoding performed by factorial-pulse encoder 2092 of FIG. 2C.

Block 12093 execution noise filling performs noise filling the result of decoding FPC. For example, block 12093 performing noise filling adds noise to frequency bands that undergo FPC decoding. In this case, the noise filling unit 12093 adds noise up to the last frequency bands in the frequency bands that undergo FPC decoding, which will be described below with reference to FIG. 27.

Decoder 12094 in FD with extension may include a low-frequency decoder 12095 in FD with extension and high-frequency decoder 12096 in FD with extension.

If the upper band Ffpc value of the frequency bands performing FPC decoding is lower than the high band Fcore value of the frequency bands performing FPC encoding, then the low-frequency decoder 12095 in the FD with extension performs encoding with the extension of the FPC decoding result and the noise filling result.

Thus, the low-frequency decoder 12095 in the FD with the extension generates frequency spectra up to the Fcore value of the upper band of the frequency bands performing FPC encoding using the frequency spectra generated by FPC decoding and noise filling.

As described above, decoded low-frequency spectra can be formed by multiplying the frequency spectra generated by the low-frequency decoder 12095 in the FD with the extension, the normalization value decoded by the normalization decoder 12091.

However, when the low-frequency decoder 12095 in the FD with the extension does not work, decoded low-frequency spectra can be formed by multiplying the frequency spectra generated by performing FPC decoding and performing noise filling by the normalization value decoded by the normalization decoder 12091.

A high-frequency decoder 12096 in FD with extension performs high-frequency decoding with extension using the results of FPC decoding and noise filling. In the current embodiment, the high-frequency decoder 12096 in the FD with the extension operates to match the encoder 2094 in the FD with the extension of FIG. 2C.

For example, an extension FD decoder 12096 in FD can inverse quantize the high-frequency energy based on the high-frequency energy information of the binary signal stream, generate an excitation signal of the high-frequency signal using the low-frequency signal in accordance with various high-frequency range extension modes, and generate a decoded high-frequency signal in accordance with using amplification so that the energy of the excitation signal can be symmetric to the inverse quantized energy. For example, various high-frequency range extension modes may include normal mode, harmonics mode, or noise mode.

In this case, the high-frequency decoder 12096 in the FD with the extension can perform energy quantization by sharing the same codebook with respect to different bit rates, as will be described in detail below with reference to FIG. 26.

If the frame to be decoded is a constant frame, then normalization decoder 12091, FPC decoder 12092, noise cancellation execution unit 12093, and FD decoder 12094 included in the FD decoder 1209 can work.

However, if the frame to be decoded is a transition frame, then the FD decoder 12094 with the extension may not work.

FIG. 12D is a block diagram of a decoding device 102 in accordance with another embodiment of the present invention.

The decoding process performed by the decoding device 102 of FIG. 12D is the inverse process to the process of FIG. 2D. Referring to FIG. 12D, decoding apparatus 102 may include a mode information checking unit 1211, an LPC decoder 1212, a CELP decoder 1213, a TD decoder 1214 with extension, an audio decoder 1215, and an FD decoder 1216 with extension.

The mode information checking unit 1211 checks the mode information of each of the frames included in the binary signal stream. The binary signal stream may be a signal corresponding to the binary signal stream in accordance with the encoding result performed by the encoding device 101 transmitted to the decoding device 102.

For example, the mode information checking unit 1211 analyzes the mode information from the binary signal stream and performs an operation of switching to one of the decoding mode with CELP or the FD decoding mode in accordance with the encoding mode of the current frame in accordance with the analysis result.

More specifically, the mode information checking unit 1211 can be switched with respect to each of the frames included in the binary signal stream, so that the frame encoded in accordance with the CELP mode can be decoded with CELP, and the frame encoded in accordance with the FD mode, can be decoded by FD.

LPC decoder 1212 performs decoding of LPC frames included in the binary signal stream.

A CELP decoder 1213 performs decoding with CELP frames encoded in accordance with the CELP mode based on the result of the check. For example, a CELP decoder 1213 decodes the adaptive and constant codebook contributions, combines the decoding results, and generates a low frequency signal corresponding to the decoded signal for the low frequency range.

Decoder 1214 in TD with the extension generates a decoded signal for the high frequency range using at least one of the result of decoding with CELP and the excitation signal of the low-frequency signal. The low-frequency excitation signal may be included in the binary signal stream. Also, the TD decoder 1208 with extension can use the LPC information decoded by the LPC decoder 1212 to generate a high frequency signal corresponding to the decoded signal for the high frequency range.

Also, in accordance with the current embodiment, the TD decoder 1214 with extension can generate a decoded signal by combining a high frequency signal with a low frequency signal generated by a CELP decoder 1214. To generate the decoded signal, the decoder 1214 in TD with the extension can additionally perform the conversion operation on the sampling frequencies of the low-frequency signal and the high-frequency signal so that they are the same.

The audio decoder 1215 performs audio decoding of the encoded audio frame based on the verification result. For example, audio decoder 1215 accesses a binary signal stream and performs decoding based on the contribution of the time domain and the contribution of the frequency domain when the contribution of the time domain is present, or performs decoding based on the contribution of the frequency domain when the contribution of the time domain is absent.

Also, audio decoder 1215 can generate a decoded low-frequency excitation signal by performing inverse frequency conversion, for example IDCT, on a signal quantized in accordance with FPC or LVQ, and generate a decoded low-frequency signal by combining the excitation signal with inverse-quantized LPC.

The decoder 1216 in the FD performs extension decoding using the result of performing audio decoding. For example, the decoder 1216 to FD converts the decoded low-frequency signal to a sampling frequency suitable for performing high-frequency decoding up-conversion, and performs frequency conversion, such as MDCT, on the converted signal. An expansion FD decoder 1216 can inverse quantize the quantized high-frequency energy, generate an excitation signal of the high-frequency signal using the low-frequency signal in accordance with various high-frequency extension modes, and generate a decoded high-frequency signal in accordance with the application of amplification so that the energy of the excitation signal can be symmetric inverse quantized energy. For example, various high-frequency range extension modes may include normal mode, harmonics mode, transient mode, or noise mode.

Also, the decoder 1216 in the FD with the extension generates a decoded signal by performing inverse frequency conversion, for example inverse MDCT, on the decoded high-frequency signal and low-frequency signal.

In addition, if the transition mode is used to expand the high frequency range, then the decoder 1216 in the FD with the extension can apply the gain calculated in the time domain so that the signal decoded after performing the inverse frequency conversion can match the decoded time envelope and combine the signal, to which the gain is applied.

Accordingly, the decoding device 102 may perform decoding of the binary signal stream based on the encoding mode of each of the frames included in the binary signal stream.

FIG. 13 is a block diagram of an extension decoder 1304 included in a decoding device 102 in accordance with an embodiment of the present invention.

Referring to FIG. 13, the extension decoder 1204 may include an inverse quantizer 1301, a gain calculator 1302, a gain application unit 1303, an artificial signal generator 1304, an envelope estimator 1305, and an envelope application unit 1306.

The inverse quantizer 1301 may inverse quantize the energy of the input signal. The process of inverse quantization of the energy of the input signal will be described in detail below with reference to FIG. fourteen.

Gain calculator 1302 can calculate the gain to be applied to the main signal based on the inverse quantized energy and the energy of the main signal. In detail, the gain can be determined by the ratio between the inverse quantized energy and the energy of the main signal. Typically, energy is determined using the sum of the squares of the amplitude of the frequency spectrum. Thus, the square root of the relationship between the inverse quantized energy and the energy of the main signal can be used.

The gain application unit 1303 can apply gain for each frequency band to determine the frequency spectrum of the SWB.

For example, gain calculation and gain application can be performed by aligning the band with the frequency band used to transmit energy, as described above. According to another embodiment of the present invention, gain calculation and gain application can be performed by dividing all frequency bands into subbands to prevent a sudden change in energy. In this case, the energies at the band boundaries can be smoothed out by interpolating the inverse quantized energies of the neighboring bands. For example, gain computation and gain application can be performed by dividing each band into three subbands, assigning the inverse quantized energy of the current band to the middle subband of three subbands of each band, and using the energy assigned to the middle band of the previous or subsequent band, and again smoothed energy by interpolation. That is, the gain can be calculated and applied in units of subbands.

This method of smoothing energy can be used as a constant type. Also, the energy smoothing method can only be applied to the necessary frames by transmitting information from the encoder 204 with an extension indicating that energy smoothing is necessary. In this case, information may be set indicating that energy smoothing is necessary, and the quantization error in all energy when energy smoothing is performed is smaller than the quantization error in all energy when energy smoothing is not performed.

The main signal may be generated using an input signal in the frequency domain. The process of generating the main signal may be performed as described below.

The artificial signal generator 1304 can generate an artificial signal corresponding to the high frequency range by copying and folding the low frequency range of the input signal in the frequency domain. The input signal in the frequency domain may be a decoded wideband (WB) signal having a sampling frequency of 32 kHz.

Envelope estimator 1305 can estimate the envelope of the main signal using a window included in the binary signal stream. This window was used by encoder 101 to estimate the envelope, and window type information can be included in the binary signal stream as a bit type and transmitted to decoding device 102.

Envelope application unit 1306 can generate a main signal by applying the estimated envelope to the artificial signal.

When the envelope estimator 602 included in the encoder 101 estimates the average frequency value for each whitening band as the envelope of the frequency belonging to the whitening band, if information indicating whether the number of frequency spectra belonging to the whitening band is large or small is transmitted to decoding device 102 , then envelope estimator 1305 in decoding device 102 may estimate the envelope based on the transmitted method. The envelope application unit 1306 can then apply the estimated envelope to the artificial signal. Otherwise, the envelope may be determined in accordance with the base coding mode used by the broadband (WB) base decoder without the need to transmit information.

The base decoder 1201 can decode the signals by classifying the signal encoding modes into a vocal encoding mode, a non-vocal encoding mode, a transient encoding mode, and a general encoding mode based on the characteristics of the signals. In this case, the envelope estimator 602 may control the number of frequency spectra belonging to the whitening band based on the decoding mode in accordance with the characteristics of the input signal. For example, if the input signal is decoded in accordance with the decoding mode with vocals, then the envelope estimator 1305 can estimate the envelope by creating three frequency spectra in the whitening band. If the input signal is decoded in a decoding mode other than the decoding mode with vocals, the envelope estimator 1305 can estimate the envelope by creating three frequency spectra in the whitening band.

FIG. 14 is a flowchart illustrating the operation of an inverse quantizer 1301 included in an extension decoder 1304 in accordance with an embodiment of the present invention.

In step S1401, the inverse quantizer 1301 may inverse quantize the selected subvector of the energy vector based on the index received from the encoder 101.

In step S1402, the inverse quantizer 1301 may inverse quantize the interpolation errors corresponding to the remaining subvectors that are not selected based on the received index.

In step S1403, inverse quantizer 1301 can calculate the remaining subvectors by interpolating the inverse quantized subvector. Then, inverse quantizer 1301 can add inverse quantized interpolation errors to the remaining subvectors. Also, inverse quantizer 1301 can calculate the inverse quantized energy by adding the average that was subtracted during the pre-processing operation, through the post-processing operation.

FIG. 15A is a flowchart illustrating an encoding method in accordance with an embodiment of the present invention.

In step S1501, the encoder 101 may lower the sampling frequency of the input signal.

In step S1502, the encoder 101 may perform basic encoding of the downsampled input signal.

In step S1503, the encoder 101 may perform frequency conversion on the input signal.

In step S1504, the encoder 101 may perform band spreading of the input signal in the frequency domain. For example, encoding device 101 may perform band extension coding using coding information determined by basic coding. In this case, the encoding information may include an encoding mode classified according to the characteristics of the input signal when basic encoding is performed.

For example, encoding device 101 may perform band extension coding as described below.

Encoding device 101 may generate a main signal of the input signal in the frequency domain using the frequency spectra of the input signal in the frequency domain. As another example, encoder 101 may generate a main input signal in the frequency domain based on the characteristics and frequency spectra of the input signal. In this case, the characteristics of the input signal can be derived through basic coding or by additional classification of the signal. Encoder 101 can estimate the energy control coefficient using the main signal. Encoder 101 can extract energy from an input signal in the frequency domain. Encoding device 101 can then adjust the extracted energy using an energy control coefficient. Encoding device 101 may quantize controlled energy.

The main signal may be generated as described below.

Encoding device 101 may generate an artificial signal corresponding to the high frequency range by copying and folding the low frequency range of the input signal in the frequency domain. Then, encoder 101 can estimate the envelope of the main signal using a window. In this case, the encoding device 101 can estimate the envelope of the main signal by selecting a window by comparing tone or correlation. For example, encoder 101 may estimate the average frequency values of each of the whitening bands as an envelope of the frequency belonging to each of the whitening bands. In detail, the encoder 101 can estimate the envelope of the main signal by adjusting the number of frequency spectra belonging to the whitening band in accordance with the basic encoding mode.

Encoding device 101 may then apply the estimated envelope to the artificial signal to form the main signal.

The energy control coefficient can be estimated as described below.

Encoder 101 can calculate the tonality of the high frequency range of the input signal in the frequency domain. Encoder 101 can calculate the key of the main signal. The encoder 101 can then calculate the energy control coefficient using the key of the high frequency range of the input signal and the key of the main signal.

Quantization of controlled energy can be performed as described below.

Encoder 101 can select and quantize a subvector and quantize the remaining subvectors using an interpolation error. In this case, the encoding device 101 may select a subvector in the same time interval.

For example, encoder 101 may perform MSVQ using at least two steps by selecting candidate subvectors. In this case, the encoding device 101 may generate an index set to minimize the MSE or WMSE for each of the candidate sub-vectors at each stage, and select the candidate sub-vector having the smallest sum of MSE or WMSE at all stages from the candidate sub-vectors. Otherwise, encoder 101 may generate an index set to minimize MSE or WMSE for each of the candidate sub-vectors at each step, restore the energy vector by inverse quantization, and select a candidate sub-vector to match the MSE or WMSE between the reconstructed energy vector and the original energy vector.

FIG. 15B is a flowchart illustrating an encoding method in accordance with another embodiment of the present invention. The encoding method of FIG. 15B may include steps that are sequentially performed by encoder 101 of one of FIG. 2A-2C. Thus, although not described herein, the above descriptions of the encoder 101 with reference to FIG. 2A-2C can also be applied to the encoding method of FIG. 15B.

In step S1505, the signal classification block 205 determines the encoding mode of the input signal based on the characteristics of the input signal.

In step S1506, if the encoding mode of the input signal is determined as CELP mode, then the CELP encoder 206 performs CELP encoding of the low-frequency signal of the input signal.

In step S1507, if the low-frequency signal of the input signal is subjected to coding with CELP, then the TD encoder 207 with extension performs encoding in TD with the extension of the high-frequency signal of the input signal.

In step S1508, if the encoding mode of the input signal is determined to be FD mode, then the frequency converter 208 performs frequency conversion on the input signal.

In step S1509, the encoder 209 in the FD performs FD encoding of the converted frequency input signal.

FIG. 15C is a flowchart illustrating an encoding method in accordance with another embodiment of the present invention. The encoding method of FIG. 15C may include steps that are sequentially performed by encoder 101 from one of FIG. 2A-2C. Thus, although not described herein, the above descriptions of the encoder 101 with reference to FIG. 2A-2C can also be applied to the encoding method of FIG. 15C.

In step S1510, the signal classification block 210 determines the encoding mode of the input signal based on the characteristics of the input signal.

In step S1511, the LPC encoder 211 extracts the LPC from the low frequency signal of the input signal and quantizes the LPC.

In step S1512, if the encoding mode of the input signal is determined to be CELP mode, then the CELP encoder 212 encodes with CELP the LPC drive signal extracted using the LPC.

In step S1513, if the LPC drive signal of the low-frequency signal of the input signal is subjected to coding with CELP, then the TD encoder 213 with extension performs coding in TD with the extension of the high-frequency signal of the input signal.

In step S1514, if the encoding mode of the input signal is determined to be FD mode, then audio encoder 214 audio encodes the LPC drive signal extracted using the LPC.

In step S1515, if the LPC drive signal of the low-frequency signal of the input signal is subjected to FD coding, then the FD encoder 215 in the FD extension performs coding in the FD with the extension of the high-frequency signal of the input signal.

FIG. 16A is a flowchart illustrating a decoding method in accordance with an embodiment of the present invention.

At step S1601, the decoding device 102 can perform basic decoding of the base-coding encoded input signal included in the binary signal stream.

At step S1602, the decoding device 102 may increase the sampling rate of the input signal decoded by the base decoding.

At step S1603, the decoding device 102 may perform frequency conversion on the input signal with an increased sampling frequency.

In step S1604, the decoding apparatus 102 may perform band widening decoding using an input signal in a frequency domain and energy information of an input signal included in a binary signal stream.

More specifically, bandwidth extension can be performed as described below.

The decoding device 102 may inverse quantize the energy of the input signal. In this case, the decoding device 101 can select and inverse quantize the subvector, interpolate the inverse quantized subvector, and add the interpolation error to the interpolated subvector, thereby quantizing the energy inversely.

Also, decoding device 102 may generate a main input signal in the frequency domain. Then, the decoding device 102 can calculate the gain to be applied to the main signal using the inverse quantized energy and the energy of the main signal. After that, the decoding device 102 can apply gain for each frequency band.

The main signal may be generated as described below.

The decoding device 102 can generate an artificial signal corresponding to the high frequency range of the input signal by copying and folding the low frequency range of the input signal in the frequency domain. Then, the decoding device 102 can estimate the envelope of the main signal using the window information included in the binary signal stream. In this case, if the window information is set to the same, then no window information is included in the binary signal stream. Thereafter, decoding device 102 may apply the estimated envelope to the artificial signal.

FIG. 16B is a flowchart illustrating a decoding method in accordance with another embodiment of the present invention. The decoding method of FIG. 16B may include steps that are sequentially performed by decoding device 102 of one of FIG. 12A-12C. Thus, although not described herein, the above descriptions of decoding apparatus 102 with reference to FIG. 12A-12C can also be applied to the decoding method of FIG. 16B.

In step S1606, the mode information checking unit 1206 checks the mode information of each of the frames included in the binary signal stream.

In step S1607, the CELP decoder 1207 performs CELP decoding of the CELP encoded frame based on the verification result.

In step S1608, the TD decoder 1208 with the extension generates a decoded high frequency band signal using at least one of a result of decoding with a CELP and a low frequency excitation signal.

In step S1609, the FD decoder 1209 performs FD decoding of the FD encoded frame based on the verification result.

The inverse frequency converter 1210 performs inverse frequency conversion on the result of performing FD decoding.

FIG. 16C is a flowchart illustrating a decoding method in accordance with another embodiment of the present invention. The decoding method of FIG. 16C may include steps that are sequentially performed by decoding device 102 of one of FIG. 12A-12C. Thus, although not described herein, the above descriptions of decoding apparatus 102 with reference to FIG. 12A-12C can also be applied to the decoding method of FIG. 16C.

In step S1611, the mode information checking unit 1211 checks the mode information of each of the frames included in the binary signal stream.

In step S1612, the LPC decoder 1212 performs decoding of the LPC frames included in the binary signal stream.

In step S1613, the CELP decoder 1213 performs decoding with the CELP frame encoded with the CELP based on the verification result.

In step S1614, the TD decoder 1214 with the extension generates a decoded high frequency band signal using at least one of the result of decoding with CELP and a low frequency excitation signal.

At step S1615, the audio decoder 1215 performs audio decoding of the encoded audio frame based on the verification result.

In step S1616, the decoder 1216 in the FD with extension performs FD decoding with the extension using the result of performing audio decoding.

Regarding other questions about encoding and decoding methods that are not described with reference to FIG. 15-16, reference should be made to FIG. 1-14.

FIG. 17 is a structural diagram of an entire structure of an encoder 1702 in accordance with another embodiment of the present invention.

Referring to FIG. 17, the encoding device 100 may include an encoding mode selector 1701 and an extension encoder 1702.

The encoding mode selector 1701 may determine the encoding mode of the band extension coding using an input signal in the frequency domain and an input signal in the time domain.

More specifically, the encoding mode selector 1701 can classify the input signal in the frequency domain using the input signal in the frequency domain and the input signal in the time domain, and determine the encoding mode of the band extension coding and the number of frequency bands in accordance with the encoding mode based on the classification result . Here, the encoding mode may be defined as a new set of encoding modes that differ from the encoding mode determined when performing the basic encoding in order to increase the performance of the encoder 1702 with the extension.

For example, coding modes can be classified into normal mode, harmonic mode, transient mode and noise mode. First, the encoding mode selector 1701 determines whether the current frame is a transition frame based on the relationship between the long-term energy of the input signal in the time domain and the energy of the high frequency range of the current frame. The transition signal portion is the portion where a sharp change in energy occurs in the time domain, and accordingly, may be the portion in which the energy of the high frequency range changes significantly.

The process of determining the other three encoding modes will now be described. First, the global energies of the previous frame and the current frame are obtained, the ratio between the global energies and the signal in the frequency domain is divided into predefined frequency bands, and then the three coding modes are determined based on the average energy and maximum energy of each of the frequency bands. Usually in harmonic mode, the difference between the maximum energy and the average signal energy in the frequency domain is the largest. In noise mode, the degree of change in signal energy is generally small. The coding modes of other signals except for two signals are defined as normal mode.

According to an embodiment of the present invention, the number of frequency bands may be determined to be sixteen in normal mode and harmonic mode, may be determined to be five in transient mode, and may be determined to be twelve in normal mode.

Encoder 1702 with the extension can select the encoding mode of the extension band coding using an input signal in the frequency domain and an input signal in the time domain. Referring to FIG. 17, the extension encoder 1702 may include a main signal generator 1703, a coefficient estimator 1704, an energy extraction module 1705, an energy controller 1706, and an energy quantizer 1707. The main signal generator 1703 and the coefficient estimator 1704 are as described above with reference to FIG. 5.

The energy recovery unit 1705 may extract energy corresponding to each of the frequency bands in accordance with the number of frequency bands determined in accordance with the coding modes. The main signal generator 1703, coefficient estimation module 1704 and energy controller 1706 may be determined for use in accordance with the encoding mode. For example, these elements may be used in normal mode and harmonic mode, but may not be used in transient mode and noise mode. The main signal generator 1703, coefficient estimation module 1704 and energy controller 1706 are as described above with reference to FIG. 5. The energy of the strip that is subject to energy regulation can be quantized by the energy quantizer 1707.

FIG. 18 is a flowchart illustrating an operation of an energy quantizer 1710 in accordance with another embodiment of the present invention.

An energy quantizer 1707 can quantize energy extracted from an input signal in accordance with a coding mode. In this case, the energy quantizer 1707 can quantize the energy of the strip that needs to be optimized for the input signal based on the amount of strip energy and the perceived characteristics of the input signal, in accordance with the encoding mode.

For example, if the encoding mode is a transient mode, then the energy quantizer 1707 with respect to the five band energies can quantize the band energy using a frequency weighting method based on the perceived characteristics of the input signal. If the encoding mode is the normal mode or harmonic mode, then the energy quantizer 1707 with respect to sixteen band energies can quantize the band energy using an unequal distribution of bits based on the perceived characteristics of the input signal. If the characteristics of the input signal are not accurate, then the energy quantizer 1707 may quantize in accordance with the general method without considering the perceived characteristics of the input signal.

FIG. 19 is a diagram illustrating an energy quantization process using an unequal discharge distribution method according to an embodiment of the present invention.

In the method of unequal distribution of bits, the perceived characteristics of the input signal, which is the purpose of the extension coding, are taken into account. Thus, relatively low frequency bands of high importance for perception can be quantized more precisely in accordance with the method of unequal distribution of discharges. To this end, the energy quantizer 1707 can classify the importance of perception by assigning the same number of bits or more bits to relatively low frequency bands compared to the number of bits assigned to other frequency bands.

For example, energy quantizer 1707 assigns more bits to relatively low frequency bands with assigned numbers from “0” to “5”. The number of bits assigned to the relatively low frequency bands with numbers from "0" to "5" may be the same. The higher the frequency band, the smaller the number of bits assigned to the frequency band by the energy quantizer 1707. Accordingly, the frequency bands assigned with numbers from “0” to “13” can be quantized, as illustrated in FIG. 19 in accordance with the distribution of bits, which is described above. Other frequency bands with assigned numbers "14" and "15" may be quantized, as illustrated in FIG. twenty.

FIG. 20 is a diagram illustrating vector quantization using intra-frame prediction in accordance with an embodiment of the present invention.

An energy quantizer 1707 predicts a characteristic value of a quantization target vector that contains at least two elements, and then can perform vector quantization of the error signal between each of the elements of the quantization target vector and the predicted characteristic value.

FIG. 20 illustrates such an intra prediction method. A method for predicting the characteristic value of the quantization target vector and deriving the error signal is as follows:

where "Env (n)" denotes the energy of the strip that is not quantized, "QEnv (n)" denotes the energy of the strip that is quantized, "p" denotes the predicted characteristic value of the target quantization vector, "e (n)" denotes the error energy. In equation (8), “e (14)” and “e (15)” are vector quantized.

FIG. 21 is a diagram illustrating an energy quantization process using a frequency weighting method in accordance with another embodiment of the present invention.

In the method of weighting frequencies, relatively low-frequency bands with a high importance for perception can be quantized more accurately by taking into account the perceived characteristics of the input signal, which is the goal of spreading coding, as in the method of unequal distribution of bits. To this end, the importance of perception is classified by assigning the same weight or more weight to relatively low frequency bands compared to weights assigned to other frequency bands.

For example, referring to FIG. 21, an energy quantizer 1707 can quantize by assigning a larger weight, for example 1.0, to relatively low frequency bands with numbers assigned from “0” to “3” and assigning a lower weight, for example 0.7, to a frequency band with assigned number “15” . To use the assigned weights, energy quantizer 1707 can calculate the optimal index using WMSE.

FIG. 22 is a diagram illustrating vector quantization in multi-stage separate and vector quantization using intra-frame prediction in accordance with an embodiment of the present invention.

Energy quantizer 1707 can perform vector quantization in normal mode, in which the number of band energies is sixteen, as illustrated in FIG. 22. Here, the energy quantizer 1707 can perform vector quantization using a method of unequal distribution of bits, intra-frame prediction, and multi-stage separate VQ with energy interpolation.

FIG. 23 is a diagram illustrating the operation of the inverse quantizer 1301 included in decoding device 102 in accordance with an embodiment of the present invention.

The operation of the inverse quantizer 1301 of FIG. 23 may be the inverse of the operation of the energy quantizer 1710 of FIG. 18. When encoding modes are used to perform encoding with the extension described above with reference to FIG. 17, the inverse quantizer 1301 may decode encoding mode information.

First, the inverse quantizer 1301 decodes encoding mode information using the received index. Then, the inverse quantizer 1301 performs inverse quantization in accordance with the decoded encoding mode information. Referring to FIG. 23, in accordance with coding modes, blocks that are targets of inverse quantization are inverse quantized in the reverse order in which quantization is performed.

A part that is quantized in accordance with a multi-stage split VQ with energy interpolation can be inversely quantized, as illustrated in FIG. 14. The inverse quantizer 1301 may perform inverse quantization using intra-frame prediction using the following equation:

where "Env (n)" denotes the energy of the strip that is not quantized, and "QEnv (n)" denotes the energy of the strip that is quantized. Also, “p” denotes the characteristic value of the target quantization vector, and “ê (n)” denotes the quantized error energy.

FIG. 24 is a block diagram of an encoder 101 in accordance with another embodiment of the present invention.

The basic operations of the elements of the encoder 101 illustrated in FIG. 24 are the same as those of the elements of the encoder 101 illustrated in FIG. 2A, except that the encoder 2404 with the extension does not receive information from the base encoder 2402 and can directly receive an input signal in the time domain.

FIG. 25 is a diagram illustrating binary signal streams in accordance with an embodiment of the present invention.

Referring to FIG. 25, the flow of binary signals 251, stream 252 and stream of binary signals the binary signals 253 correspond to N ^th frame, (N + 1) ^th frame and the (N + 2) ^th frame, respectively.

Referring to FIG. 25, the binary signal streams 251, 252, and 253 include a header 254 and a payload 255.

Header 254 may include mode information 2511, 2521, and 2531. Information 2511, 2521 and 2531 about the mode is the information about the encoding mode N ^th frame, (N + 1) ^th frame and the (N + 2) ^th frame, respectively. For example, the information 2511 about the mode is a coding mode used to encode the N ^th frame, information 2521 about the mode is a coding mode used to encode the (N + 1) ^th frame, and information 2531 about the mode is a coding mode used to encode the (N +2) ^go frame. For example, encoding modes may include at least one of a CELP mode, an FD mode, and an audio encoding mode, but the present invention is not limited to them.

Payload 255 includes basic data information in accordance with the coding modes of these frames.

For example, if N ^th frame encoded in CELP mode, payload 255 may include information 2512 and information 2513 CELP TD expansion.

In the case of an (N + 1) ^th frame encoded in FD mode, payload 255 may include FD information 2523. In the case of the (N + 2) ^th frame encoded in FD mode, payload 255 may include FD information 2532.

The payload 255 in the binary signal stream 252 corresponding to the (N + 1) ^mu frame may further include prediction data 2522. In other words, the encoding mode between adjacent frames is switched from CELP mode to FD mode, the binary signal stream 252 in accordance with the result of encoding in accordance with the FD mode may include prediction data 2522.

More specifically, as illustrated in FIG. 2B, when an encoder 101 that allows switching between the CELP mode and the FD mode performs encoding in accordance with the FD mode, a frequency conversion such as MDCT, which includes overlapping frames, is used.

Thus, if the N ^th frame and the (N + 1) ^th frame of the input signal encoded according to the CELP mode and FD respectively, the (N + 1) ^th frame can not be decoded using only coding results in accordance with the FD mode. For this reason, if the encoding mode between adjacent frames is switched from the CELP mode to the FD mode, then the binary signal stream 252 in accordance with the result of encoding in accordance with the FD mode includes prediction data 2522 representing information corresponding to the prediction.

Accordingly, the decoding side can decode the stream 252 of binary signals encoded in accordance with the mode FD, through prediction using decoded information temporary area of the current frame, for example (N + 1) ^th frame, and the result of the previous frame decoding, for example N ^th frame, based on prediction data 2522 included in the binary signal stream 252. For example, time-domain information may be overlapping time-domain spectra, but the present invention is not limited to this.

Also, the payload 255 in the binary signal stream 252 corresponding to the (N + 1) ^th frame may further include information on the previous frame mode 2524, and the payload 255 in the binary signal stream 253 corresponding to the (N + 2) ^th frame, may further include information about the previous frame mode 2533.

More specifically, the binary signal streams 252 and 253 encoded in accordance with the FD mode may further include previous frame mode information 2524 and 2533, respectively.

For example, the information 2524 about the mode of the previous frame included in a stream 252 of binary signals corresponding to the (N + 1) ^th frame may include information about the information 2511 about the mode N ^th frame, and information 2533 of the previous frame mode included in stream 253 binary signals corresponding to the (N + 2) ^th frame may include information about information 2524 about the mode of the (N + 1) ^th frame.

Thus, even if an error occurs in one of the plurality of frames, the decoding side can accurately detect a mode transition.

FIG. 26 is a diagram illustrating a method of performing frequency allocation for each frequency band in accordance with an embodiment of the present invention.

As described above, the encoder 2094 in the FD with the extension of FIG. 2C or encoder 215 in the FD with the extension of FIG. 2D can perform energy quantization by sharing the same codebook even at different bit rates. Thus, when the frequency spectrum corresponding to the input signal is divided into a predetermined number of frequency bands, an encoder 2094 in FD with extension or encoder 215 in FD with extension can assign the same frequency band to each of the frequency bands even at different transmission rates bits.

Case 261 will now be described where a frequency band of approximately 6.4 to 14.4 kHz is divided by a bit rate of 16 kbps, and case 262 where a frequency band of approximately 8 to 16 kHz is divided by a bit rate which greater than or equal to 16 kbit / s. In these cases, the frequency bandwidth of each of the frequency bands is the same even at different bit rates.

That is, the width 263 of the frequency band of the first frequency band can be equal to 0.4 kHz at a bit rate of 16 kbit / s, and the bit rate that is greater than or equal to 16 kbit / s, and the width of 264 frequency bands of the second frequency band can be equal to 0.6 kHz and at a bit rate of 16 kbit / s, and a bit rate that is greater than or equal to 16 kbit / s.

As described above, since the frequency bandwidth of each of the frequency bands is set to be the same even at different bit rates, the encoder 2094 in FD with extension or encoder 215 in FD with extension in accordance with the current embodiment can perform energy quantization by sharing the same same codebook at different bit rates.

Thus, in a configuration in which switching between CELP mode and FD mode, or between CELP mode and audio coding mode, multi-mode bandwidth expansion can be performed, and codebook sharing is performed to support different bit rates, thereby reducing size, for example , read-only memory (ROM) and simplifying implementation.

FIG. 27 is a diagram illustrating a frequency band 271 used in an encoder in an FD or a decoder in an FD in accordance with an embodiment of the present invention.

Referring to FIG. 27, frequency band 271 may be an example of a frequency band that may be used, for example, in encoder 209 in the FD of FIG. 2B and decoder 1209 in the FD of FIG. 12B in accordance with each of the tools.

More specifically, the factorial pulse encoder 2092 in the encoder 209 in the FD limits the frequency band for performing FPC encoding in accordance with the bit rate. For example, the Fcore bandwidth for performing FPC encoding may be 6.4 kHz, 8 kHz, or 9.6 kHz in accordance with a bit rate, but the present invention is not limited to this.

The factorially pulse-encoded frequency band Ffpc 272 can be determined by performing FPC in the frequency band bounded by the factorial-pulse encoder 2092. In this case, the noise filling block 12093 in the FD decoder 1209 performs noise filling in the factorial-pulse-encoded band Ffpc 272 frequencies.

In this case, if the upper band value of the factorially pulse-encoded Ffpc band 272 is lower than the upper band value of the Fcore band to perform FPC, then the low-frequency decoder 12095 in FD with extension in the decoder 1209 FD can perform low-frequency decoding with extension.

Referring to FIG. 27, low-frequency decoder 12095 in FD with spreading can perform low-frequency decoding in FD with spreading in the remaining frequency band 273 from the Fcore band except for the factorially pulse-encoded Ffpc band. However, if the Fcore frequency band is the same as the factorially pulse-coded Ffpc 272 frequency band, then low frequency decoding in FD with extension may not be performed.

A high-frequency decoder 12096 in FD with an extension in decoder 1209 in FD can perform high-frequency decoding in FD with an extension in band 274 between the high band value of the Fcore band and the high band value of the Fend frequency band according to the bit rate. For example, the upper band value of the Fend band may be 14 kHz, 14.4 kHz, or 16 kHz, but the present invention is not limited to this. Thus, using the encoding device 101 and the decoding device 102 in accordance with an embodiment of the present invention, speech and music can be efficiently encoded at different bit rates through various switching systems. Also, FD coding with extension and FD decoding with extension can be performed by sharing a codebook. Thus, it is possible to realize high-quality sound in a less complex way, even when various configurations are present. In addition, since information about the previous frame mode is included in the binary signal stream when FD coding is performed, decoding can be performed accurately even when a frame error occurs. Accordingly, using encoding device 101 and decoding device 102, it is possible to perform encoding and decoding with low complexity and low latency.

Accordingly, the voice signal and the music signal in accordance with the advanced voice service (EVS) 3GPP can be encoded and decoded in an appropriate manner.

The above methods in accordance with various embodiments of the present invention can be implemented in the form of a computer program that can be executed by various types of computing means and written to a computer-readable recording medium. A computer-readable recording medium may store program instructions, data files, data structures, or a combination thereof. Program commands may be specifically designed or created in accordance with the present invention or may be known in the field of computer software.

Despite the fact that this invention is shown and described in detail with reference to its exemplary embodiments, ordinary specialists in the art will understand that it can be made various changes in form and detail without deviating from the essence and scope of the invention, which are defined the attached claims.

Claims (20)

1. An encoding device comprising:
a signal classification unit for determining an encoding mode of an input signal based on characteristics of the input signal;
Code Excited Linear Prediction Encoder (CELP) for performing CELP coding of a low frequency signal of an input signal when an encoding mode of an input signal is defined as a CELP encoding mode;
time domain (TD) encoder with extension for performing extension coding of a high frequency signal of an input signal when a low frequency signal of an input signal is encoded with CELP;
a frequency converter for performing frequency conversion on an input signal when an encoding mode of an input signal is defined as a frequency domain (FD) mode; and
encoder in FD to perform FD encoding of the converted input signal. 2. The encoding device according to claim 1, in which the encoder in the FD contains:
a normalization encoder for extracting energy from the converted input signal for each frequency band and quantizing the extracted energy;
a factorial pulse encoder for performing factorial pulse encoding (FPC) of a value obtained by scaling a converted input signal using a quantized normalization value; and
an additional noise information generator for generating additional noise information in accordance with the implementation of the FPC,
moreover, the converted input signal input to the encoder FD is a transition frame. 3. The encoding device according to claim 1, in which the encoder in the FD contains:
a normalization encoder for extracting energy from the converted input signal for each frequency band and quantizing the extracted energy;
a factorial pulse encoder for performing factorial pulse encoding (FPC) of a value obtained by scaling a converted input signal using a quantized normalization value;
an additional noise information generator for generating additional noise information in accordance with the implementation of the FPC; and
an FD encoder with extension for performing extension coding of a high frequency signal of a converted input signal,
moreover, the converted input signal input to the encoder FD is a stationary frame. 4. The encoding device according to claim 3, in which the encoder in the FD with the extension performs energy quantization using the same codebook at different bit rates. 5. The encoding device according to claim 1, wherein the binary signal stream in accordance with the result of performing FD encoding of the converted input signal includes information about a previous frame mode. 6. An encoding device comprising:
a signal classification unit for determining a coding mode of a low-frequency signal of an input signal based on characteristics of said low-frequency signal of an input signal;
a linear prediction coefficient encoder (LPC) for extracting an LPC from said low frequency input signal and quantizing an LPC;
a code-excited linear prediction encoder (CELP) for performing CELP coding of the LPC drive signal of the low-frequency signal of the input signal extracted using the LPC when the coding mode of said low-frequency signal of the input signal is determined to be a CELP coding mode;
Time Domain Wide Encoder (TD) for performing wideband coding of a high frequency input signal when an LPC drive signal is encoded with CELP;
an audio encoder for performing audio coding of the LPC drive signal when the coding mode of the low-frequency signal of the input signal is determined as an audio mode; and
a frequency domain encoder (FD) with extension for performing extension coding of a high frequency signal of an input signal when an LPC drive signal is subjected to audio coding. 7. The encoder according to claim 6, in which the encoder in the FD with the extension performs energy quantization using the same codebook at different bit rates. 8. The encoding device according to claim 6, in which the encoder in the FD with the extension is configured to:
extract energy from the input signal;
regulate extracted energy; and
quantize adjustable energy. 9. The encoding device according to claim 8, in which the encoder in the FD with the extension is configured to extract energy corresponding to each of the frequency bands. 10. The encoding device according to claim 8, in which the encoder in the FD with the extension is configured to regulate energy using the coefficient of energy regulation, estimated from the input signal. 11. The encoding device according to claim 8, in which the encoder in the FD with the extension is configured to vectorize energy by assigning weight to the low frequency range with high perception importance. 12. The encoder according to claim 8, in which the encoder in the FD with the extension is configured to quantize energy by assigning more bits to the low frequency range with high perception importance than the high frequency range. 13. A decoding device comprising:
a mode information checking unit for checking mode information of each of the frames included in the binary signal stream;
Code Excited Linear Prediction (CELP) decoder for performing CELP decoding of a CELP encoded frame based on the verification result;
a time domain extension (TD) decoder for generating a decoded highband signal using at least one of a result of decoding with a CELP and a low frequency excitation signal;
a frequency domain (FD) decoder for performing FD decoding of an FD encoded frame based on the result of the check; and
inverse frequency converter for performing inverse frequency conversion of the result of FD decoding. 14. The decoding device according to claim 13, in which the decoder in the FD contains:
normalization decoder for performing normalization decoding based on normalization information included in the binary signal stream;
a factorial pulse coding (FPC) decoder for performing FPC decoding based on factorial pulse coding information included in a binary signal stream; and
a noise filling execution unit for performing noise filling of an FPC decoding execution result. 15. The decoding device according to claim 13, in which the decoder in the FD contains:
normalization decoder for performing normalization decoding based on normalization information included in the binary signal stream;
a factorial pulse coding (FPC) decoder for performing FPC decoding based on factorial pulse coding information included in a binary signal stream;
a noise filling execution unit for performing noise filling of an FPC decoding execution result; and
a high-frequency decoder in FD with extension for performing high-frequency decoding with extension based on the result of performing FPC decoding and the result of performing noise filling. 16. The decoding device according to claim 14, wherein the decoder in FD further comprises a low frequency encoder in FD with extension for performing coding with extension of the results of FPC decoding and noise filling when the upper band value of the frequency band performing FPC decoding is less than the upper bands at the baseband. 17. The decoding device according to claim 14, in which the high-frequency decoder in FD with the extension performs inverse quantization of energy by sharing the same codebook at different bit rates. 18. The decoding device according to claim 15, in which the decoder in the FD performs FD decoding of the frame encoded in the FD based on the information about the previous frame mode included in the binary signal stream. 19. A decoding device comprising:
a mode information checking unit for checking mode information of each of the frames included in the binary signal stream;
a linear prediction coefficient (LPC) decoder for decoding LPC frames included in the binary signal stream;
Code Excited Linear Prediction (CELP) decoder for performing CELP decoding of a CELP encoded frame based on the verification result;
a time domain (TD) decoder with extension for generating a decoded highband signal using at least one of a result of decoding with a CELP and a low frequency excitation signal;
an audio decoder for performing audio decoding of the encoded audio frame based on the verification result; and
a frequency domain (FD) decoder with extension for performing extension decoding using the result of performing audio decoding. 20. The decoding device according to claim 19, in which the decoder in the FD with the extension performs inverse quantization of energy by sharing the same codebook at different bit rates.

Images