Anthropometric data on horizontal head movements in videocommunications
Martin B�cker, Werner Blohm, and Lothar M�hlbach
- Heinrich-Hertz-Institut f�r Nachrichtentechnik
- Einsteinufer 37, D - 10587 Berlin
- Tel: +49 30 31002237, E-mail: boecker@hhi.de
Abstract
Head movement data were collected from 128 subjects during an experimental study using four different videoconferencing set-ups (factorial design of monoscopic / stereoscopic set-ups with / without motion parallax). The data include various parameters and are relevant inter alia for terminal and display designers.
Keywords
Display design, videocommunications, motion parallax, head tracking, 3D.
Introduction
In the context of a research project which investigated the potential benefits of employing HDTV- and 3DTV technologies in videoconferencing [1], horizontal head movement data of conferees were automatically collected by using a Polhemus position sensor.
This data may be of interest inter alia for:
- designers of conventional videophone terminals as a clue for defining the camera's optimal field of view,
- designers of autostereoscopic displays (i.e., stereoscopic displays that do not require to wear special glasses, see e.g. [2]) and of liquid crystal displays (LCDs), for specifying the width of viewing zones from which images should be visible as well as expected dynamic viewer behavior,
- designers of videocommunication terminals which provide motion parallax (i.e., as in natural vision, viewers see different perspectives from different positions) as indication of the range of positions, velocities and accelerations their systems should reliably work for, and
- developers of video codecs as an estimate of image content changes coding algorithms have to cope with.
The videoconferencing systems evaluated in the study included set-ups with:
- a 'standard' monoscopic image display,
- a stereoscopic image display,
- a monoscopic image display reproducing motion parallax, and
- a stereoscopic image display reproducing motion parallax.
The latter two set-ups provided dynamic images correct in perspective to give motion parallax in horizontal direction.
Method
All experimental set-ups were used for point-to-point videoconferencing with two conferees at each site. In all cases, the subjects sat at an oblong table facing the display (51" rear projection screen) at a distance of 200 cm. At a central sitting position, the distance between the heads of the two subjects was about 70 cm. Head position data were captured with 10 samples/sec using a Polhemus sensor attached to shutter glasses worn by the users.
The technical set-ups were used by 32 subjects each, i.e. 128 subjects were included in the analysis. The majority were between 20 and 30 years old. Two thirds reported an average to large conference experience and about one third went on more than one business trip per year. 15% had encountered videocommunications devices.
During the experimental sessions, each pair of subjects performed two tasks: a collaborative decision making task (an advertising strategy had to be developed), and a negotiating task (discussion about a controversial issue). Each tasks lasted about 15 minutes.
Results and Discussion
For both tasks, the statistics of acquired head movements are compiled in Tables 1 and 2, respectively.
From the statistical data presented in Tables 1 and 2, it is apparent that the liveliness of users is higher when videoconference systems provide motion parallax: greater total path lengths were covered, more direction reversals occurred, and slightly higher magnitudes of head velocities and accelerations appeared.
According to the data, maximum head accelerations between 2 and 3 m/sec2 have to be expected in videoconferences. Velocities of much more than 0.5 m/sec seem to be rather unlikely.
For all videoconference set-ups investigated, the positions taken by the subjects were on the whole within a range of (30 cm from their central sitting positions. Thus, when considering our viewing geometry (2 persons, 70 cm apart, in front of a screen / camera which is 200 cm away) a pick-up camera should have a horizontal viewing angle of about 40( in order to capture both viewers. Likewise, autostereoscopic displays should allow for stereoscopic viewing under quite a similar angle.
Images correct in perspective and with motion parallax should also be delivered within the (30 cm range of viewing positions taken by the subjects. Relating this to the apparent position of the interlocutors in our experiment (50 cm behind the screen) this gives a lateral angular region of about 30( for two viewers and smaller for one viewer respectively within which a lookaround capability should be provided by displays reproducing motion parallax.
The data presented in Tables 1 and 2 show only minor differences in head movements between the decision making task and the negotiating task.
References
- M�hlbach, L., B�cker, M., & Prussog, A. (1995). Telepresence in videocommunications: A study on stereoscopy and individual eye contact. Human Factors, 37(2), p. 290-305.
- B�rner, R. Autostereoscopic 3D-imaging by front and rear projection and on flat panel displays. Displays, 14, 1.
