Lecturer: Prof. Giampiero Salvi
This is the course I took during my exchange in KTH in Sweden.
The course note is here (it's in Chinese because I think there is sufficient resource in English on the internet 😃)
- Assignment1: Feature extraction and comparison
- Step-by-step MFCC and survey in detail through experiment
- Dynamic Time Wrapping algorithm to compare utterances
- Assignment2: Hidden Markov Model
- Forward/Backward algorithm to evaluate the likelihood of state sequence
- Viterbi algorithm to find the optimal state sequence
- Baum-Welch Algorithm to train the phoneme model
- Assignment3: Phoneme Recognition with DNN
- Use Assignment1 to extract audio feature
- Use Assignment2 to force-align phoneme as label
- Train phoneme recognticizer implemented in Deep Neural Network