Acoustic Features for Speech Recognition: From Mel-Frequency Cepstrum CoefficientsBottleNeck Features(BNF)
References (MFCC) to BottleNeck Features(BNF)
Outline
What are acoustic features?
Mel-Frequency Cepstrum Coefficients(MFCC)
MFCC (from wiki)
MFCC
Hamming Window
MFCC
Mel-Filter Bank Outputs
MFCC
Cepstral Coeffiencents
MFCC
The final step
The MFCC framework
Improvement of the MFCC frameworkImprovement of the MFCC framework
How do we let the data drive the coefficients?
Data driven transformations
Machine Learning
The Deep Neural Network
The Deep Neural Network
BottleNeck Features(BNF)

The MFCC framework

The action of applying DFT, mel-Filter bank, and DCT can be viewed as multiplying the input feature by a matrix with predefined weights.
These weights are designed by “human heuristics”