Text

Web-based hearing tests can enable collecting larger and more diverse datasets than lab-based studies. To explore the potential and limitations of data collected in online experiments, we conducted a study employing different conditions of tone-in-noise detection and speech recognition tests with the matrix sentence test. This study took an exploratory approach investigating the feasibility of large online listening experiments.

Tone-in-noise detection thresholds for broadband white noise ranging from 100 Hz to 4 kHz were measured with the tone centered at 500 Hz and 2 kHz. Detection thresholds were also measured for notched-noise conditions, in which a spectral notch was cut out around the tone frequency. In a third condition tone-detection thresholds in a pseudo-noise were measured. Here, the tone was centered around 500 Hz while the noise masker was a white noise ranging from 2 kHz to 4 kHz, effectively measuring tone-detection thresholds in quiet at 500 Hz. Speech recognition tests were performed via the Oldenburg sentence test [1] with an adaptive measurement procedure [2] to estimate the speech recognition threshold (SRT).

Data from over 700 participants from German, English, Spanish and Turkish speaking countries was collected in the online experiment and 53 participants performed identical testing in the lab, while 26 people were tested both in the lab and online to form a validation group.

Results indicate general feasibility of these tests, especially for speech in noise testing, where SRTs were found to be very similar between online and lab testing. Differences between the languages were also in line with data from the literature. However, for tone-in-noise detection thresholds some significant differences were found, especially between lab and online measurements of the same participants. These effects might be related to uncontrolled surroundings (background noise) and hence audibility issues, but needs further investigation.

References