Overcoming language barriers in speech intelligibility: Characterizing language- and talker-dependent factors and their consequences for hearing devices

• Applicant: Dr. Sabine Hochmuth
• Subject Area: Acoustics
• Term: from 2017 to 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 387101551

The analysis of speech intelligibility in noise is at the heart of auditory research and audiological applications, ranging from clinical assessment to hearing device fitting. The interaction of speaker- and language-specific factors influencing speech intelligibility in complex listening situations has not been extensively researched until now. Speech materials and speech tests are often difficult to compare across languages and are typically recorded with different speakers for different languages. The relative importance of language- and speaker-specific factors is therefore not sufficiently understood. The aim of this project is to improve comparability of speech recognition research across languages and, based on new tests, to contribute towards an improvement of hearing devices worldwide. First, this includes an investigation of speech intelligibility of different languages (German, English, Spanish) and different speakers in critical listening conditions (e.g. noise interference and reverberation) using highly comparable speech materials across languages. To this end, speech materials recorded with bilingual speakers will be used, whose pronunciation is accent-free in two languages (German/English, English/Spanish and Spanish/German). This permits to disentangle and quantitatively capture speaker- and language-specific factors influencing speech intelligibility best possible. In a second step, acoustic-phonetic correlates (e.g. vowel space area, modulation depth etc.) will be identified in the speech signal influencing speech intelligibility. In a third step, the relation between speech intelligibility and brain activity will be explored with respect to different languages and speakers by using electroencephalography (EEG). Therefore, the highly comparable speech materials and their well-controlled variation in intelligibility across speakers and languages will be used to compare cortical responses to the temporal envelope of the speech signal. The long-term goal is to find objective measures that predict the intelligibility differences across conditions and to relate them to the acoustic-phonetic differences found in the various speech materials.

• DFG Programme: Research Fellowships
• International Connection: USA
• Host: Professorin Barbara Shinn-Cunningham, Ph.D.