Evaluation of suitable parameters for describing heart rate variability with respect to auditory selective attention switches
Master Thesis of Glosauer, Iris Maria
Noise can cause stress, which leads to measurable physical reactions as well as it can impair intentional switching of auditory selective attention (ASA). In children, this can reduce the learning progress and thus disturb their development. To benefit from unobtrusive and mobile measuring methods in the investigation of intentional switching of ASA, the observability of intentional switching of ASA through measurable physical stress reactions is of interest. This way, unwanted side effects on ASA by the measuring method itself are reduced and the number of possible applications is increased. The measure for physical reactions assessed in this thesis was heart rate variability (HRV). State-of-the-art measuring of heartbeat lengths via a sensor mounted to a chest strap allowed unobtrusive and mobile data collection from 27 healthy adults (COVID-19 pandemic had prohibited the participation of children) for the determination of different HRV parameters. Simultaneously, a listening experiment, fully-within designed with repeated measures, provided different noise conditions during a categorisation task. The participants had to separate cued target stimuli from distractor stimuli from different positions within a virtual acoustic scene. The stimuli were composed from two words, which were either the from the same category or from different categories respectively. In total, there were three different levels of stimuli congruence. Reaction time and error rate served as measures for intentional switching of ASA. Linear regression was employed to assess the prediction quality of HRV with respect to intentional switching of ASA. Switching target position significantly prolonged intentionally switching of ASA, while it insignificantly increased the number of failed switches as well as the accompanying physical stress that could be observed in the HRV parameters SDNN, SDHR, SD1 and RMSSD. Further, distinguishing differing stimuli, whether semi- or totally incongruent, was barely possible. The duration of intentionally switching ASA increased for partly differing stimuli, while it decreased for totally differing stimuli. The physical stress in the body was unaffected by the congruence of stimuli. However, the parameters SDNN, SDHR, SD1, SD2, SD1/SD2 and rrHRV showed highest stress levels for target and distractor positioned to the left and right, while spatial hearing strongly simplified intentional switching of ASA. Increasing noise level decreased the duration of intentionally switching ASA, whereas the amount of failed switches was affected ambiguously. SDNN, SDHR and SD2 surprisingly yielded a stress peak for 0 dB SNR. The combination of different, partly opposing effects from noise is supposed to be the cause for these ambiguous results for different noise levels. Eventually, the best linear regression models were generated using meanHR, SDHR, SD2, SD1/SD2 and rrHRV, but none of them yielded an adequate predictive quality for intentional switching of ASA. Alternatives for the observation of intentional switching of ASA through HRV might be other machine learning methods.