9. Mediated Ear 9
トランプ大統領
"I have many friends I actually have a large group of people friends
that I have a great people but they do suffer these..."
深層学習を用いて,入力音声を特定の人物の声質に変換するソフトウェア
Imagine Cup 2017 プロダクト「NeuroVoice」
10. Mediated Ear 10
マイケルジャクソン
"I have many friends I actually have a large group of people friends
that I have a great people but they do suffer these..."
トランプ大統領
深層学習を用いて,入力音声を特定の人物の声質に変換するソフトウェア
Imagine Cup 2017 プロダクト「NeuroVoice」
11. Mediated Ear 11
マイケルジャクソン
"I have many friends I actually have a large group of people friends
that I have a great people but they do suffer these..."
トランプ大統領
深層学習を用いて,入力音声を特定の人物の声質に変換するソフトウェア
Imagine Cup 2017 プロダクト「NeuroVoice」
同時再生
30. Mediated Ear
雑音除去?話者分離?Mediated Earの違い
雑音除去サンプル
“Deep Clustering and Conventional Networks for Music Separation: Stronger Together”
http://danetapi.com/chimera
Luo, Yi, et al. "Deep clustering and conventional networks for music separation: Stronger together." Acoustics, Speech
and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017.
http://danetapi.com/chimera
? 音楽が混ざった入力音声
? フィルター結果
31. Mediated Ear
雑音除去?話者分離?Mediated Earの違い
雑音除去サンプル
“Deep Clustering and Conventional Networks for Music Separation: Stronger Together”
http://danetapi.com/chimera
Luo, Yi, et al. "Deep clustering and conventional networks for music separation: Stronger together." Acoustics, Speech
and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017.
? 音楽が混ざった入力音声
? フィルター結果
32. Mediated Ear
雑音除去?話者分離?Mediated Earの違い
雑音除去サンプル
“Deep Clustering and Conventional Networks for Music Separation: Stronger Together”
http://danetapi.com/chimera
Luo, Yi, et al. "Deep clustering and conventional networks for music separation: Stronger together." Acoustics, Speech
and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017.
http://danetapi.com/chimera
? 音楽が混ざった入力音声
? フィルター結果
? 複数話者が含まれる入力音声
? フィルター結果
33. Mediated Ear
雑音除去?話者分離?Mediated Earの違い
雑音除去サンプル
“Deep Clustering and Conventional Networks for Music Separation: Stronger Together”
http://danetapi.com/chimera
Luo, Yi, et al. "Deep clustering and conventional networks for music separation: Stronger together." Acoustics, Speech
and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017.
http://danetapi.com/chimera
? 音楽が混ざった入力音声
? フィルター結果
? 複数話者が含まれる入力音声
? フィルター結果
41. Mediated Ear
RNN / LSTMの欠点
“All class-based BLSTMs performed poorly
in non-speaker-dependent settings” [1]
[1] Hershey, John R., et al. "Deep clustering: Discriminative embeddings for segmentation and separation."
Acoustics, Speech and Signal Processing (ICASSP), 2016 IEEE International Conference on. IEEE, 2016.
LSTMを用いたDNNモデルは,
訓練データに含まれない話者同士の分離が難しい.