Yu-Ting Tsai Dylan Chen Ming Cheng Paul Fang Chao-Chih G. Huang HUAN LUNG HUANG Jimmy Liao ing chyang lin 1Tong Hwei LIN Penhao Ma George Soong Hanson Tsai Yi-chang Tu Paul Wang Tze-Chien Wang Ming Hung Wang Chih-hui Pan Shu-Nung Yao hwei-Ching Yeh Chien Hsun Yen Shih-Ho Chang Yuhsi Chen
- Yu-Ting Tsai , PhD , who's research interests:
Digital Signal Processing/Control Systems/Optimization Algorithm
- Artificial Intelligence Learning Algorithm
- Application of Electroacoustic System: hearing and measurement
- Measurement System and DSP Signal Processing
- Virtual Auditory Model
- Machine Learning Structure: supervised learning & unsupervised learning
- Sound Source Separation
1.Cocktail party effect: Ability of focus on a single source, even if the speech
environment is complex
2.Prince Shotoku Challenge : Ability of focus on multiple sources
simultaneously. The legendary Prince of Japan, Shotoku, can
listen to and understand the requests of ten people at the same time.
- The Paradigms of Musical Source Separation
- Musical Source Separation Tasks
- Mixing Models
1. Linear mixing model
2. Delayed mixing model
3. Convolutive mixing model
- System of linear equations: X=AS
Usual algebraic methods from high school: X known, A known, S unknown ; But in
source separation: unknown variables (S, sources) AND unknown coefficients
(A, mixing matrix)
- Classification : Method & Segmentation
- Multi-Class Discriminative Classification : usually performed by combining binary
classifiers
