Task dialog by native-Danish talkers in Danish and English in both quiet and noise

The zip files contain recorded conversations between 19 pairs of normal-hearing native-Danish talkers taking part in an experiment in the lab of the Hearing Systems Group at The Technical University of Denmark during October-November 2016.

Each participant pair conversed in each of four conditions:
- In their first language (Danish) without background noise     (L1-quiet)
- In their first language (Danish) with background noise        (L1-noise)
- In their second language (English) without background noise   (L2-quiet)
- In their second language (English) with background noise      (L2-noise)

The conversations were recorded in three blocks with a break between each block. Each block consisted of a conversation from each of the four conditions (in a random order). In each conversation, the pairs found 10 differences between two almost identical pictures (DiapixUK; Baker and Hazan, 2011). The pictures were randomized across conditions. The background noise was a mixed-gender 6-talker babble-like vocoded noise (ICRA 7, Dreschler et al. (2001)).

The pairs were seated in separate sound booths and could not see each other. 

Each talker wore a Shure WH20 microphone (close mic) along with a pair of Sennheiser HD650 headphones. An operator sat outside the door monitoring the conversations and had the opportunity to talk to the participants through an operator microphone. In the headphones, both talkers heard a mix of themselves, their conversational partner, and the background noise (in the noise conditions). The gains were set and held constant such that the speech level presented by the headphones was the same as the level in the soundbooth 1 m away from the talker.

In total four channels were recorded using an RME Fireface 802 soundcard at 24 bit, 48 kHz, with Matlab 2016a: 
Channel 1: the microphone of Talker 1
Channel 2: the microphone of Talker 2
Channel 3: the operator microphone
Channel 4: The mix sent to the headphones

Prior to recording, all 20 pairs provided written consent for the recordings to be used in the conversation experiment. After the recordings had been completed, we recognized that these recordings might be useful for others. Thus we approached the participants and asked if they would be willing to make the recordings publicly available. 19 out of the 20 pairs gave written consent to make the recordings publicly available. However, pair 12 asked that one of their conversations not be made publicly available (the third replicate of the conversation in Danish in quiet). Thus, this conversation and the conversations from the pair that did not provide written consent to make the recordings public (pair 9) are not included here. 

The file "Recordings_4channel_48kHz_24bit.zip" contains the original recordings. When uncompressed, the wav files are approximately 42 GB in total.

The file "Recordings_2channel_22050Hz_16bit.zip" contains only channels 1 and 2 (the microphones from both talkers) from the recordings, which have been downsampled to 22050 Hz and rescaled as 16 bit wav files. When uncompressed, the wav files are approximately 6.4 GB in total.

In both zip files, there are individual wav files for each conversation and the file names are structured as follows: TalkerPair_Condition_Replicate.wav.

For any questions, please contact 
Anna Josefine Sørensen: ajs@elektro.dtu.dk

or

Ewen MacDonald: emcd@elektro.dtu.dk

References:
R. Baker and V. Hazan. DiapixUK: task materials for the elicitation of multiple spontaneous speech dialogs. Behavior Research Methods, 43(3):761–770, 2011. ISSN 1554-3528. doi: 10.3758/s13428-011-0075-y.

W. Dreschler, H. Verschuure, C. Ludvigsen, and S. Westermann. ICRA Noises: Artificial Noise Signals with Speech-like Spectral and Temporal Properties for Hearing Instrument Assessment: Ruidos ICRA: Señates de ruido artificial con espectro similar al habla y propiedades temporales para pruebas de instrumentos auditiv. International Journal of Audiology, 40(3):148–157, 2001. ISSN 1499-2027. doi: 10.3109/00206090109073110.