[DL輪読会]SEGAN Speech Enhancement Generative Adversarial Network

•Descargar como PPTX, PDF•

1 recomendación•448 vistas

Deep Learning JP

020/02/14 Deep Learning JP: http://deeplearning.jp/seminar-2/2

SEGAN
Speech Enhancement Generative Adversarial Network
okamura masaki

目次
1.書誌事項
2.タスクの目的
3.GAN
4.提案手法(SEGAN)
5.実験結果
6.まとめ

書誌事項
year：2017
Santiago Pascual, Antonio Bonafonte, Joan Serra
- Universitat Politecnica de Catalunya,Telefonica Research(spain)
project page ：(http://veu.talp.cat/segan/)
コードも公開：(https://github.com/santi-pdp/segan)

タスクの目的
雑音下の音声をクリーンにする。
音声
雑音・騒音

GAN
データセット
(real data)
ノイズ
(乱数などから生成)
Generator
Discriminator
本物
偽物

GAN
Generator：G(x) を最小化へ Discriminator：D(x),1-D(G(z))を最大化へ
① ②

CGAN (conditional GAN)
y：追加の条件を与えるベクトル
新たな特徴を加えることが可能

LSGAN (least-suquares GAN)
学習が安定化
(a,b,c)=(-1,1,0),(0,1,1)が例として挙げられている。

提案手法(SEGAN)①
①Generator
Encoder-Decoder 構造
noisy speech
enhancement speech
②Discriminator
enhancement signal noisy signal
Discriminator
real fake

提案手法② -Generator
青：encoder
特徴を表す “c”を生み出すため
緑：decoder
(z,c)をもとに、clean speechを生成するため
損失関数
input noise signal
clean signal:

提案手法(SEGAN)③ - Discriminator
損失関数
D(x)
input noisy signal
enhancement
signal
noisy
signal
Discriminator
real fake

提案手法(SEGAN)④ - 工夫
Discriminator - 最小２乗誤差を用いて導出
(LSGANを参考)
Generator - λ=100,L1 norm (距離を表す指標)を利用

提案手法(SEGAN)④ - コードより
Discriminator loss
# TRAIN D to recognize clean audio as clean
# TRAIN D to recognize generated audio as noisy
Generator loss
# TRAIN G so that D recognizes G(z) as real
leftthomasさんのgit hub(https://github.com/leftthomas/SEGAN)からの引用

実験結果
1.Objective evaluation
PESQを除いて、性能が上がった
2.Subjective evaluation
1~5の点数をつけてもらった結果
(1が最低、5が最高)

まとめ
1.音声処理とGANの組み合わせはまだまだ増えていきそうな
ので注目していきたい。
2.自分のプロジェクトにも機械学習を取り入れていきたい。
3.貴重な発表機会を与えていただきありがとうございました。

参照
・論文(https://arxiv.org/pdf/1703.09452.pdf)
・プロジェクトページ(http://veu.talp.cat/segan/)
・ Lsgan(https://arxiv.org/pdf/1611.04076.pdf),(https://qiita.com/inoudayo/items/a98da29b735c610fd7de)
・cGAN(https://arxiv.org/pdf/1411.1784.pdf)
・PESQに関して(https://www.ntt.co.jp/qos/technology/sound/04_2.html)

Recomendados

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving PlannersDeep Learning JP

【DL輪読会】事前学習用データセットについて

【DL輪読会】事前学習用データセットについて

【DL輪読会】事前学習用データセットについてDeep Learning JP

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...Deep Learning JP

【DL輪読会】Zero-Shot Dual-Lens Super-Resolution

【DL輪読会】Zero-Shot Dual-Lens Super-Resolution

【DL輪読会】Zero-Shot Dual-Lens Super-ResolutionDeep Learning JP

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxivDeep Learning JP

【DL輪読会】マルチモーダル LLM

【DL輪読会】マルチモーダル LLM

【DL輪読会】マルチモーダル LLMDeep Learning JP

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...Deep Learning JP

【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition

【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition

【DL輪読会】AnyLoc: Towards Universal Visual Place RecognitionDeep Learning JP

Recomendados

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners

【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving PlannersDeep Learning JP

【DL輪読会】事前学習用データセットについて

【DL輪読会】事前学習用データセットについて

【DL輪読会】事前学習用データセットについてDeep Learning JP

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...

【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...Deep Learning JP

【DL輪読会】Zero-Shot Dual-Lens Super-Resolution

【DL輪読会】Zero-Shot Dual-Lens Super-Resolution

【DL輪読会】Zero-Shot Dual-Lens Super-ResolutionDeep Learning JP

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv

【DL輪読会】BloombergGPT: A Large Language Model for Finance arxivDeep Learning JP

【DL輪読会】マルチモーダル LLM

【DL輪読会】マルチモーダル LLM

【DL輪読会】マルチモーダル LLMDeep Learning JP

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...

【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...Deep Learning JP

【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition

【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition

【DL輪読会】AnyLoc: Towards Universal Visual Place RecognitionDeep Learning JP

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?Deep Learning JP

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究についてDeep Learning JP

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )Deep Learning JP

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...Deep Learning JP

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"Deep Learning JP

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "Deep Learning JP

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat ModelsDeep Learning JP

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"Deep Learning JP

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...Deep Learning JP

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...Deep Learning JP

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...Deep Learning JP

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...Deep Learning JP

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...Deep Learning JP

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...Deep Learning JP

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデルDeep Learning JP

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...Deep Learning JP

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...Deep Learning JP

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLMDeep Learning JP

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without SupervisionDeep Learning JP

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...Deep Learning JP

Más contenido relacionado

Más de Deep Learning JP

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?Deep Learning JP

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究についてDeep Learning JP

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )Deep Learning JP

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...Deep Learning JP

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"Deep Learning JP

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "Deep Learning JP

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat ModelsDeep Learning JP

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"Deep Learning JP

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...Deep Learning JP

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...Deep Learning JP

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...Deep Learning JP

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...Deep Learning JP

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...Deep Learning JP

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...Deep Learning JP

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデルDeep Learning JP

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...Deep Learning JP

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...Deep Learning JP

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLMDeep Learning JP

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without SupervisionDeep Learning JP

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...Deep Learning JP

Más de Deep Learning JP (20)

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Can Neural Network Memorization Be Localized?

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】Hopfield network　関連研究について

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】マルチモーダル基盤モデル

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】TrOCR: Transformer-based Optical Character Recognition with Pre-traine...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】HyperDiffusion: Generating Implicit Neural Fields withWeight-Space Dif...

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】大量API・ツールの扱いに特化したLLM

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】DINOv2: Learning Robust Visual Features without Supervision

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

【DL輪読会】Poisoning Language Models During Instruction Tuning Instruction Tuning...

[DL輪読会]SEGAN Speech Enhancement Generative Adversarial Network

1. SEGAN Speech Enhancement Generative Adversarial Network okamura masaki

2. 目次 1.書誌事項 2.タスクの目的 3.GAN 4.提案手法(SEGAN) 5.実験結果 6.まとめ

3. 書誌事項 year：2017 Santiago Pascual, Antonio Bonafonte, Joan Serra - Universitat Politecnica de Catalunya,Telefonica Research(spain) project page ：(http://veu.talp.cat/segan/) コードも公開：(https://github.com/santi-pdp/segan)

4. タスクの目的雑音下の音声をクリーンにする。音声雑音・騒音

5. GAN データセット (real data) ノイズ (乱数などから生成) Generator Discriminator 本物偽物

6. GAN Generator：G(x) を最小化へ Discriminator：D(x),1-D(G(z))を最大化へ ① ②

7. CGAN (conditional GAN) y：追加の条件を与えるベクトル新たな特徴を加えることが可能

8. LSGAN (least-suquares GAN) 学習が安定化 (a,b,c)=(-1,1,0),(0,1,1)が例として挙げられている。

9. 提案手法(SEGAN)① ①Generator Encoder-Decoder 構造 noisy speech enhancement speech ②Discriminator enhancement signal noisy signal Discriminator real fake

10. 提案手法② -Generator 青：encoder 特徴を表す “c”を生み出すため緑：decoder (z,c)をもとに、clean speechを生成するため損失関数 input noise signal clean signal:

11. 提案手法(SEGAN)③ - Discriminator 損失関数 D(x) input noisy signal enhancement signal noisy signal Discriminator real fake

12. 提案手法(SEGAN)④ - 工夫 Discriminator - 最小２乗誤差を用いて導出 (LSGANを参考) Generator - λ=100,L1 norm (距離を表す指標)を利用

13. 提案手法(SEGAN)④ - コードより Discriminator loss # TRAIN D to recognize clean audio as clean # TRAIN D to recognize generated audio as noisy Generator loss # TRAIN G so that D recognizes G(z) as real leftthomasさんのgit hub(https://github.com/leftthomas/SEGAN)からの引用

14. 実験結果 1.Objective evaluation PESQを除いて、性能が上がった 2.Subjective evaluation 1~5の点数をつけてもらった結果 (1が最低、5が最高)

15. まとめ 1.音声処理とGANの組み合わせはまだまだ増えていきそうなので注目していきたい。 2.自分のプロジェクトにも機械学習を取り入れていきたい。 3.貴重な発表機会を与えていただきありがとうございました。

16. 参照・論文(https://arxiv.org/pdf/1703.09452.pdf) ・プロジェクトページ(http://veu.talp.cat/segan/) ・ Lsgan(https://arxiv.org/pdf/1611.04076.pdf),(https://qiita.com/inoudayo/items/a98da29b735c610fd7de) ・cGAN(https://arxiv.org/pdf/1411.1784.pdf) ・PESQに関して(https://www.ntt.co.jp/qos/technology/sound/04_2.html)