Leveraged the success in speech data classification to extract features from time series in an unsupervised manner. In the research, we used a Symbolic Dynamic Filtering approach to unsupervised extraction of features by using "stickiness factor" and "Chinese Restaurant Process" to control disturbances-motivated transition between high level Finite States in a Dirichlet Compounded Multinomial approach and more importantly, online data analysis manner. The result below was obtained for a simulated complex Duffing system with signal-to-noise ratio = 1.
|
Real Time and Embedded Systems Control of a Twin Rotor Using MATLAB and Simulink
|
|
Plasma Turbulence Identification
A spatio-temporal analysis which used Volterra functional kernel to identify the wave-wave and wave-particle interaction mechanism and describe plasma turbulence in the bow shock vicinity of the earth. It was extended to a non-Gaussian kind of scenario. |