- Lecturer
- Shinozaki Takahiro
- Place
- Tue1-2(G224)
- Credits
- Lecture2 Exercise0 Experiment0
- Code
- 88116
- Syllabus updated
- 2015/4/6
- Lecture notes updated
- 2015/3/16
- Semester
- Spring Semester
Outline of lecture
This course provides an introduction to computer based pattern recognition systems mainly focusing on speech signal processing. The organization of the systems, machine learning techniques, search algorithms, and performance evaluation methods are described.
Purpose of lecture
The purpose of this lecture is to provide an introduction to speech recognition and machine learning. Not only fundamentals but also recent advances in the theory and practice of speech recognition techniques are explained.
Plan of lecture
Speech analysis and feature extraction
Regression and clustering
Probabilistic distributions and maximum likelihood estimation
Gaussian mixture model (GMM)
Expectation maximization (EM) algorithm
Hidden Markov model (HMM)
Word network and N-gram model
Weighted finite state transducer (WFST)
Bayesian network
Speaker adaptation
Variational Bayes
Sampling
Boltzmann machine
Multilayer perceptron
Evolutionary algorithm
Textbook and reference
Lecture notes will be handed out in class.
Reference:
* 縲碁浹螢ｰ隱崎ｭ倥す繧ｹ繝繝縲 鮖ｿ驥取ｸ螳 莉 (繧ｪ繝ｼ繝遉ｾ)
* "Pattern Recognition and Machine Learning", C. M. Bishop (Springer)
Related and/or prerequisite courses
basic understanding of linear algebra, differentials, probability and statistics
Evaluation
Midterm and final exams.
