KR20210030886A - Encoding method and decoding method for audio signal using dynamic model parameter, audio encoding apparatus and audio decoding apparatus - Google Patents

Abstract

Disclosed are an audio encoding method and audio decoding method using a dynamic model parameter, an audio encoding device, and an audio decoding device. The audio encoding method using a dynamic model parameter may use the dynamic model parameter corresponding to each of the levels when the dimension of the audio signal is reduced in an encoding network. In addition, the audio decoding method using the dynamic model parameter may use the dynamic model parameter corresponding to each of the levels when the dimension of the audio signal is extended in the encoding network.

Description

Audio encoding method and audio decoding method using dynamic model parameters, audio encoding device and audio decoding device {ENCODING METHOD AND DECODING METHOD FOR AUDIO SIGNAL USING DYNAMIC MODEL PARAMETER, AUDIO ENCODING APPARATUS AND AUDIO DECODING APPARATUS}

The present invention relates to an audio encoding method and an audio decoding method, an audio encoding device, and an audio decoding device using dynamic model parameters, and more particularly, to generating dynamic model parameters at all levels of an encoding network and a decoding network. .

Due to the recent development of deep learning technology, various deep learning technologies are being used to process voice, audio, language, and video. In particular, deep learning technology is actively used to process audio signals. When encoding or decoding an audio signal, a method for efficiently encoding or decoding an audio signal by using a neural network such as a network composed of a plurality of levels has also been proposed.

The present invention proposes a method and apparatus capable of efficiently encoding an audio signal and restoring it close to an original signal by applying an auto encoder to an audio signal in order to process the audio signal.

The present invention proposes a method and apparatus capable of effectively processing an audio signal by generating a dynamic model parameter network for each of the layers constituting an encoding network and a decoding network of an auto encoder.

An audio encoding method using a dynamic model parameter according to an embodiment of the present invention includes reducing the dimension of an audio signal through a plurality of encoding networks; Outputting a code corresponding to an audio signal of a final level whose dimension has been reduced in the encoding network; And quantizing the output code.

In the step of outputting the code, a dimension of an audio signal corresponding to an N-level layer may be reduced by using a dynamic model parameter of each of the levels of the encoding network.

The dynamic model parameter may be generated for N-1 levels for the N levels of layers.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the encoding network.

The dynamic model parameter may be determined based on a feature of a previous level to determine a feature of a next level.

The encoding network may determine a feature signal of a next level using a feature signal of a previous level and a dynamic model parameter of a next level, and the feature signal of the next level may be a signal whose dimension is reduced from that of the feature signal of the previous level. .

An audio decoding method using a dynamic model parameter according to an embodiment of the present invention includes: receiving a quantized code of an audio signal corresponding to a reduced dimension; Extending a dimension of an audio signal through a plurality of decoding networks by using the code of the audio signal; And outputting an audio signal of a final level with an expanded dimension in the decoding network, and extending the dimension of the audio signal includes N levels of the audio signal using a dynamic model parameter of each of the levels of the decoding network. The dimension of the audio signal corresponding to the layer can be extended.

The dynamic model parameter may be generated for N-1 levels for the N levels of layers.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the decoding network.

The dynamic model parameter may be determined based on a feature of a previous level to determine a feature of a next level.

The decoding network may determine a feature signal of a next level using a feature signal of a previous level and a dynamic model parameter of a next level, and the feature signal of the next level may be a signal whose dimension is extended from that of the previous level. .

An audio encoding method using a dynamic model parameter according to another embodiment of the present invention includes reducing the dimension of an audio signal through a plurality of encoding networks; Outputting a code corresponding to an audio signal of a final level whose dimension has been reduced in the encoding network; Quantizing the output code, and outputting the code comprises: dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter at some preset levels among the levels of the coding network. Can be reduced.

In the step of outputting the code, the dimension of the audio signal is reduced using dynamic model parameters for some preset levels of the coding network, and the remaining levels except for some preset levels are fixed among all levels of the coding network. The dimension of the audio signal can be reduced by using model parameters.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the encoding network.

The dynamic model parameter may be determined based on a feature of a previous level to determine a feature of a next level.

The encoding network determines a next level feature signal using a previous level feature signal and a next level dynamic or fixed model parameter, and the next level feature signal is a signal whose dimension is reduced from that of the previous level feature signal. I can.

An audio decoding method using a dynamic model parameter according to another embodiment of the present invention includes: receiving a quantized code of an audio signal corresponding to a reduced dimension; Extending a dimension of an audio signal through a plurality of decoding networks by using the code of the audio signal; And outputting an audio signal of a final level with an expanded dimension in the decoding network, and an audio signal corresponding to an N level of layers by using a dynamic model parameter at some preset levels among the levels of the decoding network. Dimensions can be expanded.

In the expanding step, the dimension of the audio signal is expanded using dynamic model parameters for some preset levels of the decoding network, and fixed model parameters for the remaining levels except for some preset levels among all levels of the decoding network. The dimension of the audio signal can be extended by using.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the decoding network.

The dynamic model parameter may be determined based on a feature of a previous level to determine a feature of a next level.

The decoding network may determine a feature signal of a next level using a feature signal of a previous level and a dynamic model parameter of a next level, and the feature signal of the next level may be a signal whose dimension is extended from that of the previous level. .

According to an embodiment of the present invention, by applying an auto encoder to an audio signal to process an audio signal, an audio signal can be efficiently encoded and restored close to an original signal.

According to an embodiment of the present invention, an audio signal can be effectively processed by generating a dynamic model parameter network for each of layers constituting an encoding network and a decoding network of an auto encoder.

1 is a diagram illustrating a detailed network of an auto encoder for encoding and decoding an audio signal according to an embodiment of the present invention.
2 is a flowchart illustrating an audio encoding method according to an embodiment of the present invention.
3 is a flowchart illustrating an audio decoding method according to an embodiment of the present invention.
4 is a diagram illustrating an audio encoding method and an audio decoding method according to an embodiment of the present invention.
5 is a diagram illustrating an audio encoding method and an audio decoding method according to another embodiment of the present invention.
6 is a diagram illustrating a process of deriving a dynamic model parameter for each layer of an auto encoder according to an embodiment of the present invention.
7 is a diagram for describing a process of deriving dynamic model parameters for a specific layer of an auto encoder according to another embodiment of the present invention.

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

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

1 is a diagram illustrating a detailed network of an auto encoder for encoding and decoding an audio signal according to an embodiment of the present invention.

In the present invention, a deep autoencoder composed of a plurality of layers may be used as a network structure for encoding or decoding an audio signal. The autoencoder has a representative deep learning structure for dimension reduction and representation learning. The deep-auto encoder may be composed of layers corresponding to each of a plurality of classes, and has a neural network that makes the input layer and the output layer the same. In the present invention, the deep autoencoder may be composed of an encoding network and a decoding network as shown in FIG. 1.

1 shows the basic structure of a deep auto encoder. In the encoding process of the deep auto encoder, the audio signal input to the encoding network is encoded in a manner that reduces the dimension compared to the input layer. Code Z is derived from the last layer of the coding network.

When the input data is high-dimensional (that is, the number of variables is large) and it is difficult to express the relationship between variables (or features), it is a low-dimensional vector that is easy to process while properly expressing the features of the data. Can be converted. Code refers to a new variable obtained by converting variables (or features) of an audio signal. In addition, the output value of the deep auto-encoder is a predicted value of the deep auto-encoder, and the decoding network is learned so that the audio signal input to the deep auto-encoder and the audio signal restored through the deep auto-encoder are identical to each other. Here, the code may be expressed as a latent variable. In addition, in the decoding process of the deep auto encoder, the audio signal is restored by extending the dimension. Layers constituting the deep auto encoder may be referred to as a hidden layer, and a hidden layer for encoding and a hidden layer for decoding may be configured to be symmetric with each other.

The audio signal input to the encoding network may be reduced in dimension through a plurality of layers constituting the encoding network of the deep auto encoder, and may be expressed as a latent vector of the hidden layer. In addition, the latent vector of the hidden layer is dimensionally extended through a plurality of layers constituting a decoding network of a deep auto encoder, so that an audio signal that is substantially the same as an audio signal input to the encoding network may be restored.

The number of layers constituting the encoding network and the number of layers constituting the decoding network may be the same or different from each other. In this case, if the number of layers constituting the encoding network and the decoding network is the same, the encoding network and the decoding network have a structure that is symmetric with each other.

According to an embodiment of the present invention, an audio signal input to an encoding network is output as a code with a reduced dimension through a plurality of layers constituting an encoding network of a deep-auto encoder. In addition, the code is reconstructed as an audio signal by extending the dimension through a plurality of layers constituting the decoding network of the deep autoencoder. In this case, a parameter for minimizing the error function of the deep auto encoder may be determined through a learning process so that the audio signal input to the encoding network and the audio signal restored through the decoding network are substantially the same.

The network structure of the deep auto encoder represents a structure such as a fully-connected (FC) or a convolutional neural network (CNN), but the present invention is not limited thereto, and any type of a plurality of layers may be used.

The meanings of the variables shown in FIG. 1 are as follows.

: 부호화 네트워크에 입력된 오디오 신호

: Audio signal input to the encoding network

: 잠재 벡터 (또는 코드, 병목)

: Latent vector (or code, bottleneck)

: 양자화된 코드

: Quantized code

: 복호화 네트워크를 통해 복원된 오디오 신호

: Audio signal restored through the decoding network

i: layer index

:

-번째 레이어의 동적 모델 파라미터

:

-Th layer dynamic model parameters

: 코드의 레이어 인덱스

: Layer index in code

: 오토인코더의 네트워크를 구성하는 전체 레이어 개수

: Total number of layers that make up the autoencoder's network

: 양자화

: Quantization

In Fig. 1, the output value of the i-th layer is

로 표현될 수 있다. 이 때, 여기서,

이고,

는 비선형 활성 함수를 나타낸다.

It can be expressed as At this time, here,

ego,

Represents a nonlinear activity function.

According to an embodiment of the present invention, the feed-forward process of an autoencoder is as follows.

Encoding process:

Quantization:

여기서,

Decryption process:

here,

And, according to an embodiment of the present invention, the learning process of the autoencoder is as follows.

Loss function:

The loss function L is an objective function and may be expressed as a weighted sum such as a Mean-Squared Error (MSE) and a bit rate that determine the encoding and decoding performance of the autoencoder. The basic purpose of an autoencoder is to make the audio signal input to the encoding network of the autoencoder almost identical to the audio signal restored through the decoding network of the autoencoder.

와 양자화 파라미터(예, 코드북)을 결정하기 위해, feed-forward 과정을 통해 복원된 오디오 신호인

과 오토인코더에 입력된 오디오 신호인

에 의해서 결정된 손실 함수를 출력 레이어에서 입력 레이어까지 역전파함(back-propagation)으로써 각각의 동적 모델 파라미터가 업데이트될 수 있다. 이를 위해, 코드(z)를 양자화하는 과정은 학습 단계에서 미분 가능한 형태로 진행되어야 한다.The dynamic model parameter of the i-th layer

And the audio signal restored through the feed-forward process to determine the quantization parameter (eg, codebook)

And the audio signal input to the autoencoder.

Each dynamic model parameter can be updated by back-propagation of the loss function determined by the output layer to the input layer. To this end, the process of quantizing the code z must be performed in a form that can be differentiated in the learning stage.

Fixed model parameters obtained through a conventional learning process frequently occur over-fitting or under-fitting to a DB for training. Simply, the fixed model parameters obtained through learning are fixed model parameters and are applied equally regardless of the characteristics of the input audio signal. Therefore, the conventional fixed model parameters are very limited in reflecting the characteristics of various input audio signals. Therefore, even if an input audio signal having a wide variety of characteristics is input, an encoding/decoding process is required in which the characteristics of the input audio signal can be well reflected.

The present invention proposes a method of dynamically deriving dynamic model parameters from layers constituting a deep auto encoder to reflect the characteristics of an audio signal based on a deep auto encoder.

2 is a flowchart illustrating an audio encoding method according to an embodiment of the present invention.

The audio encoding method of FIG. 2 may be performed by an audio encoding apparatus.

In step 201 of FIG. 2, the audio encoding apparatus may reduce the dimension of the audio signal through a plurality of encoding networks.

In step 202 of FIG. 2, the audio encoding apparatus may output a code corresponding to the audio signal of the final level whose dimension is reduced in the encoding network. Here, the audio encoding apparatus may reduce a dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter of each of the levels of the encoding network. Dynamic model parameters can be generated for N-1 levels for N levels of layers.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the encoding network. And, the dynamic model parameter may be determined based on the feature of the previous level to determine the feature of the next level. The encoding network may determine the next level feature signal by using the previous level feature signal and the next level dynamic model parameter. In this case, the feature signal of the next level is generally a signal whose dimension is reduced compared to the feature signal of the previous level.

In step 203 of FIG. 2, the audio encoding apparatus may quantize the output code.

3 is a flowchart illustrating an audio decoding method according to an embodiment of the present invention.

The audio decoding method of FIG. 3 may be performed by an audio decoding apparatus.

In step 301 of FIG. 3, the audio decoding method may receive a quantized code of an audio signal corresponding to a reduced dimension.

In step 302 of FIG. 3, the audio decoding method may extend the dimension of the feature signal through a plurality of decoding networks using quantized codes of the audio signal. The audio decoding apparatus may extend a dimension of a feature signal corresponding to an N-level layer by using a dynamic model parameter of each of the levels of the decoding network.

Here, the audio decoding apparatus may extend the dimension of the feature signal corresponding to the N levels of layers by using the dynamic model parameter of each of the levels of the decoding network. Dynamic model parameters can be generated for N-1 levels for N levels of layers.

The dynamic model parameter may be determined in a dynamic model parameter generation network independent from the decoding network. And, the dynamic model parameter may be determined based on the feature of the previous level to determine the feature of the next level. The decoding network may determine the next level feature signal by using the previous level feature signal and the next level dynamic model parameter. In this case, the feature signal of the next level is generally a signal whose dimension is extended from that of the previous level.

In step 303 of FIG. 3, the audio decoding method may output an audio signal of a final level with an expanded dimension in the decoding network.

4 is a diagram illustrating an audio encoding method and an audio decoding method according to an embodiment of the present invention.

In the case of FIG. 4, a process of generating a dynamic model parameter for all of the encoding process that is the first audio signal processing method and the decoding process that is the second audio signal processing method is shown.

As shown in FIG. 4, it is assumed that an encoding network for reducing the dimension of an audio signal in a deep-auto encoder is composed of layer b from layer 0, and a decoding network that extends the dimension is composed of layer b+1 to layer L. . Then, the encoding device may reduce the dimension of the audio signal by using the encoding network. And, the result of the reduced dimension can be quantized as a code.

In this case, according to an embodiment of the present invention, a dynamic model parameter may be generated for each layer of an encoding network. Specifically, the characteristics of the i-th layer (current layer) are determined based on the characteristics of the i-th layer (previous layer) and the dynamic model parameters of the i-th layer. In a similar manner, the features of the i+1th layer (next layer) are determined based on the features of the i+1th layer (current layer) and the i+1th dynamic model parameter. That is, the process of generating the dynamic model parameter may determine a dynamic model parameter for deriving a feature of a current layer from a feature derived from a previous layer.

In particular, in FIG. 4, a dynamic model parameter may be applied to an entire level of a network for processing a first audio signal for encoding and a second audio signal for decoding. However, the dynamic model parameter generation network can be independently applied to the encoding process and the decoding process.

Dynamic model parameters of each of the layers may be determined based on features of each of the plurality of layers. That is, according to an embodiment of the present invention, for audio encoding based on a deep-auto encoder, a model parameter of an autoencoder is dynamically calculated based on the characteristics of each of a plurality of layers constituting the autoencoder, thereby providing music and sound signals. The quality for the restoration of can be improved.

5 is a diagram illustrating an audio encoding method and an audio decoding method according to another embodiment of the present invention.

In FIG. 5, unlike FIG. 4, a dynamic model parameter may be commonly generated for an encoding process that is a first audio signal processing method and a decoding process that is a second audio signal processing method. That is, the dynamic model parameter generation network can be applied in common to each other for an encoding process and a decoding process. Like FIG. 4, a dynamic model parameter may be generated for each level for encoding an audio signal, and a dynamic model parameter may be generated for each level for decoding an audio signal.

6 is a diagram for describing a process of deriving a dynamic model parameter for each layer of an auto encoder according to an embodiment of the present invention.

Referring to FIG. 6, an encoding network corresponding to an encoding device and a decoding network corresponding to a decoding device are illustrated. The encoding network may be composed of b+1 layers, and the decoding network may be composed of L-b-1 layers.

Referring to FIG. 6, a dynamic model parameter generation network corresponding to each of layers constituting an encoding network and a decoding network may be defined as Gi. Here, i means the index of the layer. As for model generation parameters, FC, CNN, etc. can be used in deep-learning form.

로부터 현재 레이어 i의 특징

를 계산하기 위해 필요한 동적 모델 파라미터

가 결정될 수 있다.Referring to FIG. 6, characteristics of a previous layer i-1 for layers constituting an encoding network and a decoding network including an input layer

From the features of the current layer i

Dynamic model parameters required to calculate

Can be determined.

In Fig. 6, the coding network encodes and quantizes the input signal, and the decoding network decodes the quantized result. In this case, the dynamic model parameters used in the encoding network and the decoding network may be determined through a separate network (dynamic model parameter generation network) according to characteristics of each of the layers constituting the encoding network and the decoding network.

는 레이어 1의 특징을 추출하기 위해 사용된다.As illustrated in FIG. 6, a dynamic model parameter of a layer may be determined from a dynamic model parameter generation network for each of all layers constituting an encoding network and a decoding network. In FIG. 6, dynamic model parameters determined in layer 0

Is used to extract layer 1 features.

7 is a diagram for describing a process of deriving dynamic model parameters for a specific layer of an auto encoder according to another embodiment of the present invention.

As shown in FIG. 7, in addition to generating dynamic model parameters for all layers of an encoding network and a decoding network to reduce complexity, dynamic model parameters may be generated only for a specific layer.

는 이전 레이어(i=0, L-1)의 특징에 따라 각각 동적으로 결정될 수 있다. 다만, 부호화 네트워크의 입력 레이어와 복호화 네트워크의 출력 레이어 이외에 다른 레이어에 대한 모델 파라미터

은 고정 모델 파라미터를 사용할 수 있다.In FIG. 7, a dynamic model parameter of a layer 1 of an encoding network and a dynamic model parameter of a layer L of an output layer of a decoding network

May be determined dynamically according to the characteristics of the previous layer (i=0, L-1). However, model parameters for layers other than the input layer of the encoding network and the output layer of the decoding network

Can use fixed model parameters.

In FIG. 7, layers 2 to L-1 are set as a fixed network, but the present invention is not limited thereto. Layers included in the fixed network covering the coding and decoding networks may be determined differently according to a loss function including a signal quality and a compression rate according to coding and decoding and a quantization method.

Consequently, according to an embodiment of the present invention, by deriving dynamic model parameters for layers constituting an encoding network and a decoding network, it is possible to expect an effect of improving signal reconstruction quality and compression rate according to encoding and decoding.

Meanwhile, the method according to the present invention is written as a program that can be executed on a computer and can be implemented in various recording media, such as a magnetic storage medium, an optical reading medium, and a digital storage medium.

Implementations of the various techniques described herein may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or combinations thereof. Implementations include a data processing device, e.g., a programmable processor, a computer, or a computer program product, i.e. an information carrier, e.g., machine-readable storage, for processing by or controlling the operation of a number of computers. It may be implemented as a computer program tangibly embodied in an apparatus (computer readable medium) or a radio signal. Computer programs such as the above-described computer program(s) may be recorded in any type of programming language, including compiled or interpreted languages, and as a standalone program or in a module, component, subroutine, or computing environment. It can be deployed in any form, including as other units suitable for the use of. A computer program can be deployed to be processed on one computer or multiple computers at one site or to be distributed across multiple sites and interconnected by a communication network.

Processors suitable for processing a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. In general, the processor will receive instructions and data from read-only memory or random access memory or both. Elements of the computer may include at least one processor that executes instructions and one or more memory devices that store instructions and data. In general, a computer may include one or more mass storage devices that store data, such as magnetic, magnetic-optical disks, or optical disks, or receive data from or transmit data to them, or both. It may be combined to be Information carriers suitable for embodying computer program instructions and data are, for example, semiconductor memory devices, for example magnetic media such as hard disks, floppy disks and magnetic tapes, Compact Disk Read Only Memory (CD-ROM). ), Optical Media such as DVD (Digital Video Disk), Magnetic-Optical Media such as Floptical Disk, ROM (Read Only Memory), RAM (RAM) , Random Access Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), and the like. The processor and memory may be supplemented by or included in a special purpose logic circuit structure.

Further, the computer-readable medium may be any available medium that can be accessed by a computer, and may include both a computer storage medium and a transmission medium.

While this specification includes details of a number of specific implementations, these should not be construed as limiting to the scope of any invention or claimable, but rather as a description of features that may be peculiar to a particular embodiment of a particular invention. It must be understood. Certain features described herein in the context of separate embodiments may be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination. Furthermore, although features operate in a particular combination and may be initially described as so claimed, one or more features from a claimed combination may in some cases be excluded from the combination, and the claimed combination may be a sub-combination. Or sub-combination variations.

Likewise, although operations are depicted in the drawings in a specific order, it should not be understood that such operations must be performed in that particular order or sequential order shown, or that all illustrated operations must be performed in order to obtain a desired result. In certain cases, multitasking and parallel processing can be advantageous. In addition, separation of the various device components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and the program components and devices described are generally integrated together into a single software product or packaged in multiple software products. It should be understood that you can.

On the other hand, the embodiments of the present invention disclosed in the specification and drawings are merely presented specific examples to aid understanding, and are not intended to limit the scope of the present invention. It is apparent to those of ordinary skill in the art that other modified examples based on the technical idea of the present invention may be implemented in addition to the embodiments disclosed herein.

Claims (20)

In the audio encoding method using dynamic model parameters,
Reducing the dimension of the audio signal through a plurality of encoding networks;
Outputting a code corresponding to an audio signal of a final level whose dimension has been reduced in the encoding network;
Quantizing the output code
Including,
The step of outputting the code,
An audio encoding method for reducing a dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter of each of the levels of the encoding network. The method of claim 1,
The dynamic model parameter,
An audio encoding method generated for N-1 levels for the N levels of layers. The method of claim 1,
The dynamic model parameter,
An audio encoding method determined by a dynamic model parameter generation network independent from the encoding network. The method of claim 1,
The dynamic model parameter,
An audio encoding method that is determined based on the features of the previous level to determine the features of the next level. The method of claim 1,
The encoding network,
The feature signal of the next level is determined using the feature signal of the previous level and the dynamic model parameter of the next level,
The audio encoding method wherein the next level feature signal is a signal whose dimension is reduced from that of the previous level feature signal. In the audio decoding method using dynamic model parameters,
Receiving a quantized code of the audio signal corresponding to the reduced dimension;
Extending a dimension of an audio signal through a plurality of decoding networks by using the code of the audio signal;
Outputting an audio signal of a final level with an expanded dimension in the decoding network
Including,
Extending the dimension of the audio signal,
An audio decoding method for extending a dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter of each of the levels of the decoding network. The method of claim 6,
The dynamic model parameter,
An audio decoding method generated for N-1 levels for the N levels of layers. The method of claim 6,
The dynamic model parameter,
An audio decoding method determined in a dynamic model parameter generation network independent from the decoding network. The method of claim 6,
The dynamic model parameter,
An audio decoding method that is determined based on the features of the previous level to determine the features of the next level. The method of claim 6,
The decryption network,
The feature signal of the next level is determined using the feature signal of the previous level and the dynamic model parameter of the next level,
The audio decoding method wherein the next level feature signal is a signal whose dimension is extended from that of the previous level feature signal. In the audio encoding method using dynamic model parameters,
Reducing the dimension of the audio signal through a plurality of encoding networks;
Outputting a code corresponding to an audio signal of a final level whose dimension has been reduced in the encoding network;
Quantizing the output code
Including,
The step of outputting the code,
An audio encoding method for reducing a dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter in some of the levels of the encoding network. The method of claim 11,
The step of outputting the code,
For some preset levels of the encoding network, the dimension of the audio signal is reduced by using a dynamic model parameter,
An audio encoding method for reducing a dimension of an audio signal using a fixed model parameter for the remaining levels of the entire levels of the encoding network except for some preset levels. The method of claim 11,
The dynamic model parameter,
An audio encoding method determined by a dynamic model parameter generation network independent from the encoding network. The method of claim 11,
The dynamic model parameter,
An audio encoding method that is determined based on the features of the previous level to determine the features of the next level. The method of claim 11,
The encoding network,
The audio signal of the next level is determined using the audio signal of the previous level and the dynamic model parameter of the next level,
The audio signal of the next level is a signal whose dimension is reduced from that of the audio signal of the previous level. In the audio decoding method using dynamic model parameters,
Receiving a quantized code of the audio signal corresponding to the reduced dimension;
Extending a dimension of an audio signal through a plurality of decoding networks by using the code of the audio signal;
Outputting an audio signal of a final level with an expanded dimension in the decoding network
Including,
An audio decoding method for extending a dimension of an audio signal corresponding to an N-level layer by using a dynamic model parameter in some of the levels of the decoding network. The method of claim 16,
The expanding step,
For some preset levels of the decoding network, the dimension of the audio signal is extended by using a dynamic model parameter,
An audio decoding method for extending a dimension of an audio signal using a fixed parameter for the remaining levels of all levels of the decoding network except for some preset levels. The method of claim 16,
The dynamic model parameter,
An audio decoding method determined in a dynamic model parameter generation network independent from the decoding network. The method of claim 16,
The dynamic model parameter,
An audio decoding method that is determined based on the features of the previous level to determine the features of the next level. The method of claim 16,
The decryption network,
The feature signal of the next level is determined using the feature signal of the previous level and the dynamic or fixed model parameter of the next level,
The audio decoding method wherein the next level feature signal is a signal whose dimension is extended from that of the previous level feature signal.

Images