1. Speaker Classification through Deep Learning
Jacob Morris
Alex Douglass
Luke Woodbury 1

2. Overview Goals Potential Applications Learn more about deep learning!
Create a neural network that will classify voice recordings based on gender, age, natural language, etc.
Potential Applications
Research
Security 2

3. Software Dependencies
Python 2.7
Keras 1.2.2
Theano
Matplotlib 3

4. Hardware GeForce TitanX (Pascal) 12gb memory 4 4

5. Speech Accent Archive WAV files Categorizations
2300+ different speakers
All recorded speaking same paragraph
Categorizations
Age
Gender
English Residence
Natural Language
Country
Learning Style
Etc. 5

6. The Essence of Deep Learning

7. Artificial Neural Networks (ANN) 7

8. Recurrent Networks Layer "remembers" data 8 8

9. LSTM Long Short Term Memory 9 9

10. Problem Type Sequence Classification Supervised Learning
Assign classification label(s) to input sequences
Supervised Learning
Each training sample includes the correct output for that sample 10

11. Variations of Model Topologies
Inputs
Sequence of amplitudes
Discrete Fourier transform of the segment
Hidden Layers
Variable
Outputs
Any subset of data categories 11

12. Training Challenges Process of Exploration Many parameters to tune
Results vague, must be interpreted
Days required to train a new model 12

13. Terminology Sample Batch Epoch Base unit of training data
1/100 of a second of audio
Batch
Group of samples
4 seconds of consecutive samples
Epoch
Number of batches required to train on entire training data set
In our case, 2310 batches

14. Terminology Sample Batch Epoch Base unit of training data
1/100 of a second of audio
Batch
Group of samples
4 seconds of consecutive samples
Epoch
Number of batches required to train on entire training data set
In our case, 2310 batches

15. Loss Measure of how close an output signal is to its expected value
Categorical Cross Entropy
Emphasizes correct answer

16. Learning Rate
Determines how big of adjustments to make for given loss values

17. Accuracy
Considered correct if the expected output neuron’s activation value is the greatest among all neurons for that category

18. Initial Attempts Features Short sample lengths WAV inputs only
Trained on training set of only 2 speakers 18

19. Results 19

20. False Hope Features Changes Short sample lengths
Trained on training set of only 2 speakers
Changes
Both input types 20

21. Results 21

22. Hope Features Changes Short sample lengths Both input types
Trained on training set of only 2 speakers
Changes
Train on single batch per speaker per pass through training set
Reduced learning rate 22

23. Results 23

24. Confirmation Features Changes Short sample lengths Both input types
Train on single batch per speaker per pass through training set
Changes
Trained on full training set of speakers 24

25. Results 25

26. Refinement Features Changes Short sample lengths Both input types
Train on single batch per speaker per pass through training set
Changes
True Validation
Decaying learning rate
Epoch duration increased 26

27. Results 27

28. UI 28

29. Conclusion 29

30. Works Cited
Weinberger, Steven. (2015). Speech Accent Archive. George Mason University. Retrieved from 30