SlideShare a Scribd company logo
1 of 26
Bilinear map of filter‐bank outputs 
for DNN‐based speech recognition
Tetsuji Ogawa1, Kenshiro Ueda1, Kouichi Katsurada2, 
Tetsunori Kobayashi1, Tsuneo Nitta1,2
1Waseda University
2Toyohashi University of Technology
1
Abstract
Aim
• Accurate representation of DNN inputs
Approach
• Quadratic expansion (QE) of acoustic features
Effectiveness
2
FBANK FBANK‐QE FBANK + FBANK‐QE
18.8 17.3 15.8
Phoneme Error Rate (%) on TIMIT (dev)
Quadratic Expansion of Features
3
x1x1 x1x2 x1x3 x1x4
x2x1 x2x2 x2x3 x2x4
x3x1 x3x2 x3x3 x3x4
x4x1 x4x2 x4x3 x4x4
x1
x2
x3
x4
x1x1
x2x1
x3x1
x3x3
x4x3
x4x4
D
x 
   xxxg  tri
vec
tri
vec
vectorization
  2/)1( 
 DD
xgx  x

quadratic 
expansion
Quadratic expansion makes it possible to extract 
a more precise structure of TF pattern. 
4
xx 
QE of FBANK
x
FBANK
xt xt+1xt-1
/b/
/d/
/g/
/b/
/d/
/g/
(a) FBANK (b) QE of FBANK
xt xt+1xt-1
xt xt+1xt-1
xt xt+1xt-1
6
QE can improve class separability.
Tensor Feature Extraction
7
Context
window
(+5/‐5)
LDA QE
(2)
ty
FBANK
feature
(3)
tz
Compressed
FBANK feature
(4)
tv
Tensor 
feature
 tt zz tri
vec
bases:
tensor space
bases:
 ia  ji aa 
ti ya 
(1)
FBANK
sequence
tx
23 253 40 820
DNN System Architecture
8
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
# of hidden units: 1024
# of hidden layers: 5
# of BN units: 40
# of outputs: 1951 (states)
# of hidden units: 1024
# of hidden layers: 5
# of outputs: 1951 (states)
[M. Karafiat et al., ``BUT ASR system for Babel surprise evaluation 2014,’’ Proc. SLT2014, pp.501‐506, 2014.]
9
FBANK
x11
FBx11
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
FBANK
xN
QELDA
FBxN‐QE
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
Phoneme Recognition on TIMIT
• Speech materials: 
• Training: 3696 utterances
• Development: 400 utterances
• Test: 192 utterances
• Language model: Phoneme bigram
10
D
FBANK
xN
QELDA 0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
• N: context window size
• D: Dimensionality reduced by LDA
1. Effect of Dimension Reduction for 
Quadratic Expansion
11
D
FBANK
x5
QELDA
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
FBANK
x5
QE
12
TIMIT dev
13
Compression of FBANK can help in 
FBANK‐QE‐based DNN system.
TIMIT dev
14
TIMIT dev
2. Effectiveness of Quadratic Expansion
15
40
FBANK
xN
QELDA
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
FBANK
x5
16
TIMIT dev
17
TIMIT dev
18
TIMIT dev
19
TIMIT test
20
3. Effect of System Combination
40
FBANK
x5
QELDA
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
FBANK
x5
0
5
10
‐5
‐10
context
x21
(‐10/+10)
BN Feature Extractor Classifier
+ lattice‐level combination 
using MBR decoding
21
Individual System
System Combination
TIMIT test
[FBANK] [MFCC][FBANK] [MFCC]
+QE
+QE
[FBANK‐QE]
[MFCC‐QE]
22
Individual System
System Combination[FBANK] [MFCC]
TIMIT test
[FBANK] + [MFCC]
[FBANK] [MFCC]
+QE
+QE
[FBANK‐QE]
[MFCC‐QE]
23
[FBANK] + [FBANK‐QE]
[FBANK‐QE] + [MFCC‐QE]
[MFCC] + [MFCC‐QE]
Individual System
System Combination
[FBANK] + [MFCC]
[FBANK] [MFCC]
+QE
TIMIT test
+QE
+QE
[FBANK‐QE]
[MFCC‐QE]
24
Use of Speaker Adapted Features
Individual System
System Combination
TIMIT test
[MFCC+fMLLR]
[MFCC+fMLLR‐QE]
[MFCC+fMLLR] + [MFCC+fMLLR‐QE]
+QE
25
[MFCC+fMLLR]
[MFCC+fMLLR‐QE]
[MFCC+fMLLR] + [MFCC+fMLLR‐QE]
[MFCC+fMLLR‐QE] + [FBANK‐QE] + [MFCC‐QE]
[FBANK‐QE]
[MFCC‐QE]
+QE
Individual System
System Combination
Use of Speaker Adapted Features
TIMIT test
Conclusion
26[FBANK‐QE] + [MFCC‐QE]
[FBANK] [MFCC]
[MFCC‐QE]
+QE
+QE
• QE of acoustic features is taken as input of DNN.
• QE yields improvement over original acoustic features.
+QE
[FBANK‐QE]
[FBANK] + [MFCC]

More Related Content

Similar to Bilinear map of filter-bank outputs for DNN-based speech recognition (7)

GTC Taiwan 2017 GPU 平台上導入深度學習於半導體產業之 EDA 應用
GTC Taiwan 2017 GPU 平台上導入深度學習於半導體產業之 EDA 應用GTC Taiwan 2017 GPU 平台上導入深度學習於半導體產業之 EDA 應用
GTC Taiwan 2017 GPU 平台上導入深度學習於半導體產業之 EDA 應用
 
Filter-Type Fault Detection and Exclusion (FDE) on Multi-Frequency GNSS Receiver
Filter-Type Fault Detection and Exclusion (FDE) on Multi-Frequency GNSS ReceiverFilter-Type Fault Detection and Exclusion (FDE) on Multi-Frequency GNSS Receiver
Filter-Type Fault Detection and Exclusion (FDE) on Multi-Frequency GNSS Receiver
 
Solvedproblems 120406031331-phpapp01
Solvedproblems 120406031331-phpapp01Solvedproblems 120406031331-phpapp01
Solvedproblems 120406031331-phpapp01
 
Dsp&a
Dsp&aDsp&a
Dsp&a
 
Design Method of Directional GenLOT with Trend Vanishing Moments
Design Method of Directional GenLOT with Trend Vanishing MomentsDesign Method of Directional GenLOT with Trend Vanishing Moments
Design Method of Directional GenLOT with Trend Vanishing Moments
 
Ff tand matlab-wanjun huang
Ff tand matlab-wanjun huangFf tand matlab-wanjun huang
Ff tand matlab-wanjun huang
 
Ff tand matlab-wanjun huang
Ff tand matlab-wanjun huangFf tand matlab-wanjun huang
Ff tand matlab-wanjun huang
 

More from pcl-lab

More from pcl-lab (13)

分布類似度に基づく健全性指標と風車異常検知システムの早期運用における効果
分布類似度に基づく健全性指標と風車異常検知システムの早期運用における効果分布類似度に基づく健全性指標と風車異常検知システムの早期運用における効果
分布類似度に基づく健全性指標と風車異常検知システムの早期運用における効果
 
あらゆる風車に適用可能な状態監視技術を目指して~風車主要機器におけるデータ駆動型異常検知とその評価~
あらゆる風車に適用可能な状態監視技術を目指して~風車主要機器におけるデータ駆動型異常検知とその評価~あらゆる風車に適用可能な状態監視技術を目指して~風車主要機器におけるデータ駆動型異常検知とその評価~
あらゆる風車に適用可能な状態監視技術を目指して~風車主要機器におけるデータ駆動型異常検知とその評価~
 
画像情報を用いた黒毛和牛種の乗駕行動の検知に関する検討
画像情報を用いた黒毛和牛種の乗駕行動の検知に関する検討画像情報を用いた黒毛和牛種の乗駕行動の検知に関する検討
画像情報を用いた黒毛和牛種の乗駕行動の検知に関する検討
 
漁獲量における心理尺度と漁獲量予測器の最適化への利用
漁獲量における心理尺度と漁獲量予測器の最適化への利用漁獲量における心理尺度と漁獲量予測器の最適化への利用
漁獲量における心理尺度と漁獲量予測器の最適化への利用
 
画像情報による黒毛和牛種の状態識別に基づいた分娩予兆検知システム
画像情報による黒毛和牛種の状態識別に基づいた分娩予兆検知システム画像情報による黒毛和牛種の状態識別に基づいた分娩予兆検知システム
画像情報による黒毛和牛種の状態識別に基づいた分娩予兆検知システム
 
映像情報による肉牛の分娩検知システムにおけるクラウドソーシングを用いた誤検出抑制
映像情報による肉牛の分娩検知システムにおけるクラウドソーシングを用いた誤検出抑制映像情報による肉牛の分娩検知システムにおけるクラウドソーシングを用いた誤検出抑制
映像情報による肉牛の分娩検知システムにおけるクラウドソーシングを用いた誤検出抑制
 
畳み込みニューラルネットワークに基づく風車異常検知システムにおける判断根拠の可視化に関する検討
畳み込みニューラルネットワークに基づく風車異常検知システムにおける判断根拠の可視化に関する検討畳み込みニューラルネットワークに基づく風車異常検知システムにおける判断根拠の可視化に関する検討
畳み込みニューラルネットワークに基づく風車異常検知システムにおける判断根拠の可視化に関する検討
 
正常稼働状態の表現学習に基づく風車異常検知
正常稼働状態の表現学習に基づく風車異常検知正常稼働状態の表現学習に基づく風車異常検知
正常稼働状態の表現学習に基づく風車異常検知
 
Tandem connectionist anomaly detection: Use of faulty vibration signals in fe...
Tandem connectionist anomaly detection: Use of faulty vibration signals in fe...Tandem connectionist anomaly detection: Use of faulty vibration signals in fe...
Tandem connectionist anomaly detection: Use of faulty vibration signals in fe...
 
映像情報を用いた分娩時の牛の状態推定
映像情報を用いた分娩時の牛の状態推定映像情報を用いた分娩時の牛の状態推定
映像情報を用いた分娩時の牛の状態推定
 
定置網漁における漁獲過程モデルを用いたシロサケの日単位漁獲量予測
定置網漁における漁獲過程モデルを用いたシロサケの日単位漁獲量予測定置網漁における漁獲過程モデルを用いたシロサケの日単位漁獲量予測
定置網漁における漁獲過程モデルを用いたシロサケの日単位漁獲量予測
 
Adaptive training of vibration-based anomaly detector for wind turbine condit...
Adaptive training of vibration-based anomaly detector for wind turbine condit...Adaptive training of vibration-based anomaly detector for wind turbine condit...
Adaptive training of vibration-based anomaly detector for wind turbine condit...
 
正常・損傷の表現学習に基づく風力発電システム異常検知技術の高度化
正常・損傷の表現学習に基づく風力発電システム異常検知技術の高度化正常・損傷の表現学習に基づく風力発電システム異常検知技術の高度化
正常・損傷の表現学習に基づく風力発電システム異常検知技術の高度化
 

Recently uploaded

Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
Christo Ananth
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
ssuser89054b
 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
dharasingh5698
 
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 

Recently uploaded (20)

Unleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leapUnleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leap
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.ppt
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
 
chapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringchapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineering
 
Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
Call for Papers - African Journal of Biological Sciences, E-ISSN: 2663-2187, ...
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
 
Online banking management system project.pdf
Online banking management system project.pdfOnline banking management system project.pdf
Online banking management system project.pdf
 
University management System project report..pdf
University management System project report..pdfUniversity management System project report..pdf
University management System project report..pdf
 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
 
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete RecordCCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
 
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdf
 
Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . ppt
 
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
 
Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
 
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptxBSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
 

Bilinear map of filter-bank outputs for DNN-based speech recognition