Cross-Corpus Speech Emotion Recognition Based on Multiple Kernel Learning of Joint Sample and Feature Matching

Table 1

Acoustic features and statistical functionals.

Raw acoustic features + deltas	Statistical functional
Pitch	Mean, stand deviation, kurtosis
Root mean square energy	Skewness, minimum, maximum
Zero cross rate	Relative position  .range
Harmonic to noise ratio Mel-frequency; Cepstral coefficients mean square error of; (1–12 MFCC)	Two liner regression coefficients; Mean square error of; liner regression