The overall objective of this research by the National Criminal Technology Research, Test, and Evaluation Center was to conduct experimental quantitative and qualitative research to determine whether quantitative metrics for video quality in the video teleconferencing (VTC) for pretrial-release hearings could be found that correlate highly with human subjects' perceived video quality.
The study team identified reference videos that would test conditions that might influence the VTC system's ability to capture digitally and display scenes reasonably found in a courtroom. Such scenes might influence viewer's perception of demeanor, including variation in contrast between subject and the background, initial white balance, angle of the lighting on the subject, distance or angle between camera and subject, and reflective surfaces. A total of 138 clips representing various levels of introduced jitter and packet loss were created. The study found that participant viewers believed they were able to discern adequately facial expressions of the subject in the video despite noticeable levels of jitter and packet loss distortions. Thus, there was usefulness in videos with some levels of apparent noise; therefore, subjective rating of the utility of VTC should be further explored to determine a threshold of acceptability for automated video quality assessment tools. Although more study of this issue is needed, objective measures are apparently more conservative than human participants in the scoring of videos. Consequently, if the objective measure determines that a video is bad or poor, there is a high likelihood the video is not usable; however, if the objective measure determines that the video is fair, good, or excellent, it is likely the video is usable for pretrial release hearings. 27 tables, 6 figures, and 38 bibliographic listings
Report (Grant Sponsored)
Date Published: December 1, 2018