Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
406559 | Neurocomputing | 2014 | 10 Pages |
This paper proposes the application of complexity measures based on nonlinear dynamics for emotional speech characterization. Measures such as mutual information, dimension correlation, entropy correlation, Shannon entropy, Lempel–Ziv complexity and Hurst exponent are extracted from the samples of three databases of emotional speech. Then, statistics such as mean, standard deviation, skewness and kurtosis are applied on the extracted measures. Experiments were conducted on the Polish emotional speech database, on the Berlin emotional speech database and on the LCD emotional database for a three-class problem (neutral, fear and anger emotional states). A procedure for feature selection is proposed based on an affinity analysis of the features. This feature selection procedure is accomplished to select a reduced number of features over the Polish emotional database. Finally, the selected features are evaluated in the Berlin emotional speech database and in the LDC emotional database using a neural network classifier in order to assess the usefulness of the selected features. Global success rates of 72.28%, 75.4% and 80.75%, were obtained for the Polish emotional speech database, the Berlin emotional speech database and the LDC emotional speech database respectively.