Subband-based Independent Component Analysis

Objective

The goal of the subband-based Independent Component Analysis (ICA) is to recover independent sources given only sensor observations that are unknown mixtures of the unobserved source signals and noise.

Motivation

ICA has many applications in speech enhancement and recognition, telecommunications, biomedical signal analysis, and image enhancement and recognition. Current ICA algorithms often fail in the presence of strong noise and even when successful are slow.

Subband-based Independent Component Analysis

A new algorithm, the subband-based ICA algorithm, was proposed to separate independent signals by Yuan Qi, P.S. Krishnaprasad, and S. Shamma in 1999. The simulation results demonstrated that the algorithm is very robust and fast in the presence of noise, and suitable in dynamic situation.

Applications in the Separation of Mixed Speech Signals

Simulation results.

Video Analysis

Current Research Interests:
1. ICA for Feature Extraction and Object Recognition
2. Support Vector Machine , Sparse Coding and ICA
3. EM, Factor Analysis, and HMM
4. Bayesian Network and Turbo Codes

Test Image for Face Detection:
talkshow
hollywood1
News
Man