Comparative Performance Analysis of Algorithms for Emotions Estimation

Algorithms for estimation speech emotions using speech analysis techniques are presented in this paper. The primary focus of the algorithms is to analyze the path of the fundamental frequency F
0
(t) to accurately determine emotional states. The algorithms are trained in the same conditions. During the training phase, the algorithms are trained using test signals that contain 60% of the base signals. Decision lines were created by analyzing the parameters of the fundamental frequency in the fields (F
0
, σ
²
) and (F
0
, T). Algorithm performance evaluation is performed using the confusion matrix. A comparative analysis was also performed, comparing the results obtained using the SVM algorithm and the CDT algorithm used in the paper, for the assessment of different emotional states such as Normal/Anger, Normal/Boredom and Normal/Anxiety. To show the effectiveness of the algorithm, the results are presented in tables and graphs.