Navigation auf uzh.ch

Suche

Department of Computational Linguistics

Speaker Individuality in Phonetics and Speech Sciences: Speech Technology and Forensic Applications

Research in language sciences typically focuses on the analysis of linguistic systems and on aspects of language variation determined by sociolinguistic parameters such as the geographical and social origin of the speaker or the situational context. Research in speech sciences, on the other hand, is mainly concerned with the dynamics of speech production and perception, so that interspeaker acoustic variability – far from being a disturbance factor or ‘noise’ – may itself constitute an object of inquiry. Moreover, the speaker-specific dimensions of variation are of great relevance in several domains of applied sciences such as speech technologies (e.g., automatic speech and speaker recognition) or forensic phonetics.

This conference provides a forum for discussing and sharing research experiences related to different aspects of speaker individuality which are investigated from a multi-disciplinary perspective. Among the possible research topics we suggest the following:

  • Speaker-specific acoustic properties of vowel and consonants (e.g., formants, spectral moments or VOT)
  • Speaker-specific acoustic properties of intonational patterns (e.g., the range and median of F0 or particular contours)
  • Temporal aspects which contribute to speaker individuality (e.g., articulation and speech rate, rhythm metrics)
  • Individual aspects of voice quality (such as creaky or breathy voice) and their acoustic measurements (e.g., harmonics to noise ratio)
  • Aspects related to the geographical origin of individual speakers (phonetic features of dialects or regional varieties of standard languages)
  • Socio-phonetic aspects of speech variation including gender, age, class or ‘community of practice’
  • Effects of voice disguise on speaker individuality and recognition
  • Building and using corpora and databases for forensic purposes
  • Qualitative and quantitative requirements for acoustic stimuli used as forensic evidence
  • Audio enhancement techniques for improving speaker recognition in noisy signals
  • Statistical models and tools for the analysis of speaker individuality and speaker recognition
 

Following the tradition of AISV conferences, the programme will also include open sessions on any aspect of speech science research.