Oscillatory EEG activity – a valid measure of aided listening effort in noise
The recognition of a single speaker in a multi-speaker environment is a difficult listening task. To solve that kind of task one has to spend effort, in the following referred to as listening effort (LE). The gain and compression of the hearing instruments (HI) are thought to affect the LE required by the hearing impaired listener in such listening situations. However, there is a lack of a suitable, standardized method to quantify LE.
Consequently, the first goal of this work was to verify the recently proposed objective measure of LE by using electroencephalography (EEG) in a unified framework utilizing the EEG phase-reset hypothesis (Strauss et al., 2013; Bernarding et al., 2014) in a demanding, as well as realistic listening environment. The second goal was to utilize the proposed measure to observe the effect of different HI settings on the exerted LE. To do so, the impact of four different HI settings and two different listening task difficulties (LTD) on the LE of thirty hearing impaired subjects was observed in a selective, real-speech listening task – the Numbers in Babble Paradigm (NIB). NIB task consists of an auditory number comparison task introduced by Wöstmann et al. (2015) that has been embedded into an eight-speaker free-field cafeteria-noise environment. As a result, NIB offers a well-controlled but nonetheless ecologically valid selective listening situation. The participants have to perform an auditory number comparison task masked by a distracting talker while they are sitting in a multi-speaker environment.
HI setting A, B and C all had an adaptive compression with static characteristic, but differed in gain and compression. Setting D had an adaptive and therefore situation-dependent compression. For the quantification of LE the on-going oscillatory EEG activity has been recorded. Based on those recordings, the proposed objective measure was calculated, i.e. the Objective Listening Effort based on oscillatory EEG data (OLEosc). For comparison, the subjects also performed a subjective LE rating on a seven-point scale. Additionally, response time, decision accuracy and decision confidence were recorded as well. The results show that the OLEosc correlates with LTD, as well as with the response time, decision accuracy and processing time, in all tested conditions. These findings lead us to the assumption that the applied objective measure is a good indicator for LE. Furthermore, the results also suggest that OLEosc might be more sensitive to small variances in LE than the subjective LE rating scale.