1. AGH, NTIA, Intel, IRCCyN, Opticom, Technicolor

2.  Wide Range of Quality ◦ Original—Heavily Compressed  VGA  60 clips  10 SRC ◦ 50% English Language ◦ 50% Music Only  ACR 5-level  Subjective test used light compression

3.  Country  Native language  Audio presentation device ◦ Speakers / earbuds / headphones  Monitor size (7” to 42”)  Viewing distance, angle (8º to 20º)  Screen brightness, color (calibrated, default)  Lighting (20 lux to 200 lux)  Controlled laboratory / public area

4.  10 datasets  6 Labs ◦ 6 Controlled Environments ◦ 4 Public Environments

5. 1. NTIA 2. NTIA public 3. Intel 4. IRCCyN 5. IRCCyN public 6. Technicolor 7. Technicolor public 8. AGH 9. AGH public 10. OPTICOM

6. 12345678910 11.000.950.980.97 0.980.970.960.95 2 1.000.950.94 0.930.960.940.93 30.980.951.000.98 0.990.980.97 4 0.940.981.000.980.960.97 0.96 50.970.940.98 1.000.960.970.960.970.96 60.970.930.980.96 1.000.990.97 0.95 70.980.960.990.97 0.991.000.97 0.96 80.970.940.980.970.960.97 1.000.96 9 0.930.970.960.97 0.961.000.97 100.950.930.970.96 0.950.96 0.971.00 #2893425 24 151415

9. 1. NTIA 2. NTIA public 3. Intel 4. IRCCyN 5. IRCCyN public 6. Technicolor 7. Technicolor public 8. AGH 9. AGH public 10. OPTICOM

10.  CI drops as number of subjects increases  1/sqrt(N)

11.  Blue Controlled  Red Public  Controlled  Similar Performance  Public  Need More Subjects ◦ 24 in Controlled ◦ 35 in Public

12. 13467 1 14%15%16%13% 314% 12%11%10% 415%12% 16%13% 616%11%16% 10% 713%10%13%10%  Percent of Trials  One dataset distinguished between PVS pairs; and the other dataset did not  Opposite conclusions rare < 0.03%, 8 subjects

13.  Native language / speech comprehension  Culture / country of origin  Lighting  Background noise  Wall color  Objects on the wall  Viewing distance  Monitor calibration  Color blindness  Vision good but not 20/20  Hearing (not tested)  Translation of ACR scale labels

14.  Most Important Variables 1.Total Number of Subjects 2.People—opinions differ among subjects  MOS appears relative ◦ Order of impairments consistent ◦ Use same subjects when comparing factors  Environment Constraints Significantly Less Impact on MOS than Differences Between People