The Phonetics Cluster is concerned with human spoken language, more specifically the study of speech production, speech perception, and speech synthesis.
The study of speech production is concerned with the analysis of the acoustic signal as well as of articulatory movements, for example, of the tongue or lips. Among other things, it investigates how the sounds of a language or a variety are generated, how they differ from other sounds, languages or varieties and how their productions vary between speakers. Another factor that is analyzed is voice quality to assess whether certain speakers or speaker group groups have an especially creaky or breathy voice.
The study of speech perception is concerned with how listeners "pick out" individual sounds from the acoustic speech signal and combine them into words. A further subject of investigation is which acoustic properties are particularly important for listeners in order to recognize certain sounds and subsequently words, but also which acoustic properties allow listeners to recognize speakers as a speaker of a certain variety, for example.
Speech synthesis is concerned with the artificial simulation of the human voice. In order to create a model voice, acoustic recordings are analyzed and processed by machine learning algorithms. Our research develops new methods for speech synthesis of language varieties and evaluates the results with regard to intelligibility and perceived naturalness. In addition, speech synthesis is used in research on human speech production and perception as well as in bioacoustics.
The methods used in phonetics are manifold. Data is collected in the laboratory and in the field with informants both at work and at home. Depending on the object of study, measurements of speech production are carried out using sound recordings (for acoustic analysis) or ultrasound tongue imaging (for articulatory analysis). Video recordings, face tracking and electroglottographic (EGG) measurements (e.g. for voice analysis) are also used. In studies on speech perception, participants are presented with natural or synthesized speech samples (e.g. sounds, syllables, words, or sentences) and are asked to indicate which sound or word they hear, how “well” a word is pronounced and whether it is an actual word, etc. Both their answers and reaction times are then measured. Measurements of eye movements, that is to say what kind of speakers and/or images the listeners are looking at while listening to words or sentences, can provide information about how the speech signal is perceived in real time.
The aim of our research is to understand how human language works in order to test and expand theoretical models of language production and perception as well as linguistic theories. The research focus of the group is "Austrian language diversity in production and perception." Research topics include language change, both historical and on-going, language development in first as well as foreign languages, and human-machine interaction.