Objective and Subjective Evaluation of an Expressive Speech Corpus

Benchmarking classification models for emotion recognition in natural speech: A multi-corporal study

“…Ensemble techniques have also been widely used in the field of emotion recognition and are often compared to single classifiers [4], [23], [28], [33], [37], [40]. Our future work will consider their comparison with single classifiers and determining the most suited architecture for an ensemble that performs emotional speech recognition.…”

“…These techniques have often been compared on emotional speech assets [4], [23], [28], [33], [37], [40], but not on more than two natural/elicited datasets or for statistically significant differences. Each of these classifiers was used in 24% of 84 papers we have surveyed in average.…”

“…Different researchers proposes values from k = 1 [23], [30] to k = 20 [14]. Usually the value of k is found by trial and error, different values are taken and the performance of these classifiers is compared [16], [17].…”

“…The leaf nodes represent the classifications. The most commonly used algorithm for constructing decision trees in emotion recognition from speech is C4.5 [11], [23], [28], [30], [37].…”

“…1. The structure of an MLP [5] frequently used SVM kernel function in the domain of emotion recognition in speech is the radial basis function (RBF) kernel [11], [28], [35], [42], although some researchers also use the polynomial kernel [23], [28], [31] or the linear kernel [2], [26], [41]. In this study we use the linear and RBF kernels, excluding the polynomial kernel because of its slow training speed.…”