Dataset of Audiovisual Speech for AR Telepresence Studies (Speech Recordings)

Dataset of speech recordings made in the anechoic chamber "Lampio" at Aalto University.

21 Participant ("P1" - "P21")

Four parts are included:

1) Conversations: Ten different scripted three-part conversations ("C1" - "C10").
Each participant is in two of the conversations.
Each of the three parts is recorded by all three participants in that conversation ("S1" - "S3"). 
See assignment_conversations.xlsx for details

2) Harvard_Sets: Sets 25 and 36 of the Harvard sentence lists

3) Sentence 1 from List 25 in five different voice levels (from "barely not whispering" to "screaming as loud as you can")

4) Native_Language: List 25 translated to native languages of 12 of the participants

(French, Finnish, Hebrew, Hindi, Spanish (Mexico), Spanish (Chile), Catalan, Latvian, Italian, Polish, Romanian, German)

Each file contains data from three receivers:

Ch 1: GRAS 40 HF 1" low-noise meausurement microphone. 1.5 m away from the subject
Ch 2: RØDE NT1 large diaphragm condenser microphone. 2 m away from the subject
Ch 3: DPA 4060. Attached to the subject's clothes

Calibration.wav: Calibration data for Ch1. (Recorded using a B&K 4231 Calibrator 1kHz, 94 dB) 

Accompanying video data can be obtained by personal request from nils.meyer-kahlen@aalto.fi