Fraunhofer IDMT, HSA
Abstract:
Conversations with one or more other people are an important part of everyday life. However, communication can become very difficult when there is background noise or even competing talkers, such as in a restaurant. The resulting reduction of speech intelligibility leads to a lack of information. At the same time, the listening effort increases leading to a higher level of fatigue due to sustained effortful listening. Therefore, it is of interest to evaluate the resulting fatigue objectively. For this purpose, this study measures the Pupillary Unrest Index (PUI), which is based on the pupil diameter’s fluctuation over time. From a pilot study, we concluded that the PUI is significantly influenced by daily fatigue resulting from a working day, enabling its use in further investigations. Our main study aims to determine how fatigue is affected by a speaker extraction algorithm applied to a competitive talker scenario. The processing intends more favorable listening conditions and thereby less task-induced fatigue. Hence, speech recognition tasks with interfering talkers were carried out twice, separated into two sessions: once with speech material preprocessed by the algorithm and once without. In addition to the PUI, different speech tests were employed as well as subjective scaling methods.