Your browser does not support the video tag. Please use IE9+ or Google Chrome.

00:09
1. Slide 100:02
2. Minimum-Classification-Error (MCE) and Discriminative Training02:01
3. Slide 102:09
4. Minimum-Classification-Error (MCE) and Discriminative Training02:02
5. Minimum-Classification-Error (MCE)00:26
6. Minimum-Classification-Error (MCE) and Discriminative Training00:04
7. Minimum-Classification-Error (MCE)03:10
8. Minimum-Classification-Error (MCE) and Discriminative Training00:02
9. Minimum-Classification-Error (MCE)01:04
10. Minimum-Classification-Error (MCE) Training00:12
11. Minimum-Classification-Error (MCE)00:02
12. Minimum-Classification-Error (MCE) and Discriminative Training00:01
13. Minimum-Classification-Error (MCE)01:58
14. Minimum-Classification-Error (MCE) Training00:28
15. Minimum-Classification-Error (MCE)00:49
16. Minimum-Classification-Error (MCE) Training00:57
17. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:23
18. Minimum-Classification-Error (MCE) Training00:14
19. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:27
20. Minimum-Classification-Error (MCE) Training00:19
21. Minimum-Classification-Error (MCE)00:29
22. Minimum-Classification-Error (MCE) Training00:02
23. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:17
24. Minimum-Classification-Error (MCE) Training00:01
25. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:11
26. Minimum-Classification-Error (MCE) Training00:01
27. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:06
28. Minimum-Classification-Error (MCE) Training00:11
29. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point01:10
30. Minimum-Classification-Error (MCE) Training01:21
31. Gradient Descent Algorithm00:33
32. Minimum-Classification-Error (MCE) Training00:10
33. Gradient Descent Algorithm01:14
34. Minimum-Classification-Error (MCE) Training00:25
35. Gradient Descent Algorithm00:41
36. Minimum-Classification-Error (MCE) Training00:01
37. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:06
38. Minimum-Classification-Error (MCE) Training00:08
39. Minimum-Classification-Error (MCE)00:25
40. Minimum-Classification-Error (MCE) Training00:01
41. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:46
42. Minimum-Classification-Error (MCE) Training00:02
43. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:03
44. Minimum-Classification-Error (MCE) Training00:25
45. Minimum-Classification-Error (MCE)00:03
46. Minimum-Classification-Error (MCE) Training00:01
47. 1 𝑑 = 1 1+𝑒𝑥𝑝 −𝛾(𝑑−θ) l(d) →0 when d →-∞l(d) →1 when d →∞θ: switching from 0 to 1 near θγ: determining the slope at switching point00:03
48. Minimum-Classification-Error (MCE) Training00:01
49. Gradient Descent Algorithm01:24
50. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition01:32
51. LatticePhone Accuracy00:33
52. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition00:37
53. LatticePhone Accuracy01:04
54. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition00:01
55. Gradient Descent Algorithm00:01
56. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition00:14
57. LatticePhone Accuracy01:52
58. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition00:04
59. LatticePhone Accuracy02:26
60. Discriminative Training and Minimum Phone Error Rate (MPE) Training For Large Vocabulary Speech Recognition00:01
61. LatticePhone Accuracy01:15
62. “ Minimum Classification Error Rate Methods for Speech Recognition”, IEEE Trans. Speech and Audio Processing, May 1997“Segmental Minimum Bayes-Rick Decoding for Automatic Speech Recognition”, IEEE Trans. Speech and Audio Processing, 2004“Minimum Phone Error and I-smoothing for Improved Discriminative Training”, International Conference on Acoustics, Speech and Signal Processing, 2002“Discriminative Training for Automatic Speech Recognition”, IEEE Signal Processing Magazine, Nov 201201:20
63. Subspace Gaussian Mixture Model03:29
64. Subspace Gaussian Mixture Model03:45
65. Subspace Gaussian Mixture Model00:04
66. Subspace Gaussian Mixture Model00:12
67. Subspace Gaussian Mixture Model01:10
68. Subspace Gaussian Mixture Model00:15
69. Subspace Gaussian Mixture Model00:01
70. Subspace Gaussian Mixture Model00:05
71. Subspace Gaussian Mixture Model00:01
72. Subspace Gaussian Mixture Model01:13
73. Subspace Gaussian Mixture Model00:30
74. References for Subspace Gaussian Mixture Model02:12
75. Neural Network — Classification Task01:30
76. Neural Network — 2D Feature Space00:35
77. Neural Network ‒ Multi-Dimensional Feature Space00:48
78. Neural Network — Neurons01:46
79. Neural Network01:41
80. Neural Network Training – Back Propagation00:06
81. Gradient Descent Algorithm00:11
82. Neural Network Training – Back Propagation00:07
83. Gradient Descent Algorithm00:05
84. Neural Network Training – Back Propagation00:05
85. Neural Network00:02
86. Neural Network Training – Back Propagation00:29
87. Gradient Descent Algorithm01:28
88. Gradient Descent Algorithm02:38
89. Neural Network — Formal Formulation00:29
90. References for Neural Network02:12
91. Spectrogram00:13
92. Spectrogram00:39
93. Gabor Features (1/2)00:34
94. Gabor Features (2/2)00:21
95. Gabor Features (1/2)00:20
96. Gabor Features (2/2)00:31
97. Gabor Features (1/2)00:31
98. Gabor Features (2/2)01:42
99. Integrating HMM with Neural Networks00:38
100. Gabor Features (2/2)