Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Two-Staged Acoustic Modeling Adaption for Robust Speech Recognition by the Example of German Oral History Interviews

: Gref, Michael; Schmidt, Christoph Andreas; Behnke, Sven; Köhler, Joachim


Institute of Electrical and Electronics Engineers -IEEE-:
IEEE International Conference on Multimedia and Expo, ICME 2019. Proceedings : 8-12 July 2019, Shanghai, China
Piscataway, NJ: IEEE, 2019
ISBN: 978-1-5386-9552-4
ISBN: 978-1-5386-9553-1
International Conference on Multimedia and Expo (ICME) <2019, Shanghai>
Forschungsinfrastrukturen für die Geistes- und qualitativen Sozialwissenschaften; 01UG1511B; KA3 Kölner Zentrum für Analyse und Archivierung audiovisueller Daten
Conference Paper
Fraunhofer IAIS ()
acoustic signal processing; learning (artificial intelligence); speech recognition; spontaneous speech; german oral history interview; acoustic modeling adaption; robust speech recognition; annotated speech; reverberation data augmentation; automatic speech recognition system; acoustic recording condition; elderly people speech; word error rate; transfer learning; training; training data; history; data model; reverberation; adaptation model; robust speech recognition; domain adaption; transfer learning; multi-condition training; data augmentation; oral history

In automatic speech recognition, often little training data is available for specific challenging tasks, but training of state-of-the-art automatic speech recognition systems requires large amounts of annotated speech. To address this issue, we propose a two-staged approach to acoustic modeling that combines noise and reverberation data augmentation with transfer learning to robustly address challenges such as difficult acoustic recording conditions, spontaneous speech, and speech of elderly people. We evaluate our approach using the example of German oral history interviews, where a relative average reduction of the word error rate by 19.3% is achieved.