Automatic Discrimination of Emotion from Spoken Finnish

Abstract

In this paper, experiments on the automatic discrimination of basic emotions from spoken Finnish are described. For the purpose of the study, a large emotional speech corpus of Finnish was collected; 14 professional actors acted as speakers, and simulated four primary emotions when reading out a semantically neutral text. More than 40 prosodic features were derived and automatically computed from the speech samples. Two application scenarios were tested: the first scenario was speaker-independent for a small domain of speakers while the second scenario was completely speaker-independent. Human listening experiments were conducted to assess the perceptual adequacy of the emotional speech samples. Statistical classification experiments indicated that, with the optimal combination of prosodic feature vectors, automatic emotion discrimination performance close to human emotion recognition ability was achievable.

On enabling techniques for personal audio content management

“…Used setup is shortly summarized in this section together with a short description of the emotion classification procedure and the results. The algorithmic details and the performance figures for emotion detection have been published in [6] and [15] and are not presented here in detail due to space limitations. Based on the brief results, however, conclusions on the overall performance of the system can be drawn.…”

“…The classifiers were tested using a leave-one-out cross-validation method to maximize the utilization of statistical data. For detailed algorithmic description and a full description of the test results with our emotional data subset the interested reader should confer [6] and [15]. Inside The classifier setup was trained also for gender detection purpose.…”

“…The basic classification set required in audio/video content analysis includes the classes silence, speech, music, and environmental sounds [1]. Subsequently for example speaker segmentation and clustering [2], [3], [4], [5], as well as gender and speaker emotion detection [6] can be performed on speech category data.…”

Pervasive Social Computing: augmenting five facets of human intelligence

“…Since everybody's creature character is unique, and is usually stable over a fairly long period, this personal quality is different from others, and is difficult to imitate (Tabatabaei et al 2008). Speech-based interaction has been applied in healthcare services (e.g., electronic medical records (Wang et al 2003)), military programs for highperformance fighter aircraft (e.g., F-16) (Speech_rec-ognition 2011), and emotion recognition (Jones and Jonsson 2005;Toivanen et al 2004), etc. Widely used algorithms are statistically based, such as Hidden Markov models (HMMs) (Mana and Pianesi 2006) and dynamic time warping (DTW) (Rabiner and Juang 1993).…”

ExProsodia

“…[ANG et al, s.d. ;FUJISAWA et al, 2003;TOIVANEN et al, 2004;VOGT et al, 2005;COOK et al, 2006;VIDRASCU;DEVILLERS, 2007;RONG et al, 2007;NEIBERG;ELENIUS, 2008;BUSSO et al, 2009;YANG;LUGGER, 2010;LAUKKA et al, 2011).…”