Tag: feature-extraction
-
Modulation-domain feature enhancement
These algorithms work with sequences of feature vectors and analyze or modify the time trajectories of acoustic features from one short-time frame to the next. Citation: J. Pohjalainen, P. Alku: Multi-scale modulation filtering in automatic detection of emotions in telephone speech, in Proc. ICASSP, Florence, Italy, May 4-9, 2014. GitHub link
-
Robust spectrum analysis by linear predictive coding
These are all-pole filter models whose coefficients are estimated from short segments (frames) of audio signal and can represent the spectrum envelope. Common to these variants is non-uniform weighting of the information contained within the analysis frame according to some criterion. Citations: Please see this page, section “Robust spectrum analysis”. GitHub link
-
Feature selection code
Matlab/Octave implementations of feature selection methods for machine learning applications, suited also for higher dimensionalities. Citation for all methods: J. Pohjalainen, O. Räsänen and S. Kadioglu,“Feature Selection Methods and Their Combinations in High-Dimensional Classification of Speaker Likability, Intelligibility and Personality Traits”,Computer Speech and Language, 29(1), pp. 145-171, 2015 For RSFS specifically: O. Räsänen, J. Pohjalainen, […]