Creation and utilisation of the MediaTeam Emotional Speech Corpus

The MediaTeam Emotional Speech Corpus is currently the largest database of emotional speech for colloquial modern Finnish, containing simulated emotional content. The specific aim of the research is to investigate in detail the phonetic and phonological/linguistic correlates of basic or primary emotions in spoken Finnish, to develop statistical classification methods of emotional speech signals and to develop methods to utilise this knowledge in speech corpus search engines. The vocal expression of emotion in speech is approached utilising both the abstract phonological information on the type of pitch contour and the distribution of focal sentence accents and the concrete phonetic information on the actual implementation of F0 curves and other global (continuously variable) prosodic parameters. The annotation procedure reflects these two levels, the aim being to investigate both the linguistic and paralinguistic ways of the vocal expression of emotion. The algorithms developed for prosodic parameter estimation, the statistical classification of emotional speech, and the estimation of the degree of emotions can open up new opportunities for future Internet database technologies and methods for content-based information retrieval from audio databases.