Semester: 10th Semester, Master Thesis Title

8.1.Test phase one

The procedure on which this test was conducted was to create a test website located at keepitsimpleidiot.com/test which prompted visitors to enter their age and gender as the only personal information test subjects had to disclose. This information was collected in case there should be a large discrepancy on the ratings. If such a discrepancy would exist the groups would be divided between male and females. This would disclose if males and or females rated the portrayed facial expressions differently. Therefore the age of the test participants was also asked, to be able to divide the age groups accordingly if age had an influence on smile ratings. A mail was sent to relatives and friends with a short instruction set on what was required of the test participant. It was kept short and to the point, in order to avoid people deterring from the test due to time. The test participants were tasked with rating the level of smile on a scale of 1 to 10, 1 being no smile and 10 being a big smile. There was no time restraint on the pictures, but once a rating on a specific picture had been given, the next picture would show without the ability to go back and change the rating. This was done to avoid people second-guessing their evaluation based on the forth-coming picture.

8.1.1.Test Results – Test Subjects

Test Subjects: General: Average age: 39. Highest age: 63. Youngest age: 11. Male test subjects: Number of males: 22, average age: 36. Female test subjects: Number of females: 19, average age: 42. Figure 15 shows each picture used in the test with the average smile rating by the combined male and female test groups (The spread sheet from test phase one, test subjects, can be found in Appendix 16.1)


Figure - Test Pictures from Test Phase One in order of appearance with the smile rating for each picture denoted below the pictures

Each rating given by the test subjects will in chapter 9.2 assist the computer in determining the correct smile rating for the pictures.

8.1.2.Test Results – Computer

Figure 16 shows the calculated smile ratings given by the computer. The last column – Weighting Scale – shows the ratings after the dimension scale has been calculated. The weighting scale is the smile assessment of each picture by the computer bottom (The spread sheet from test phase one, computer results, can be found in Appendix 16.2)


Figure - Test Phase One - Computer Results

Before calculating the smile rating by the computer, the approach in obtaining the results from the computer will be analysed. The distance between the centre of the nose and the mouth was calculated three times, the resulting average is then used as the level of smile in each picture. The approach of using the average was to leverage the uncertainty in how the program calculates the region of interest. As the calculation is dependent on the user marking the location of the eyes of the test subjects, the location of the markers can vary from each calculation – as can be seen in the differences from Run1, Run2 and Run3. Therefore the average was calculated and used as the rating of the smile. Furthermore as picture 3, 7, 11, 18, 23 and 26 only contain the mouth they will not be included in the analysis by the computer, as the computer cannot estimate the level of smile due to the missing upper half of the face.



Figure - Picture 24 from Test Phase One

Lastly, the pictures that were chosen for test phase one were selected before the programming of the algorithm had begun. The pictures were selected as they were found to cover a wide range of smiles, both positive and negative displays of such. The only requirements for the pictures where a clear shot of the face, the angle or lighting conditions did not matter. Unfortunately the algorithm utilised was found to have detection problems when certain areas of the face were obstructed. Figure 16 is of a picture used in test phase one; the picture displays a female with her hands around her cheeks. Due to the hands the algorithm could not determine the correct region of interest, i.e. the algorithm could not correctly specify the location of the mouth, the eyes, nose and facial boarders. As can be seen in Table 1&2, the graph shows the comparison between the results calculated by the computer and the ratings given by the test subjects in test phase one. On both graphs, on the x-axis, picture 19 is of said female covering her face; the computer rated the smile as high whereas the test subjects rated the smile as low. The fault in the algorithm was found by comparing the results from the computer and the test subjects. Therefore pictures originally used in test phase one will have to be omitted from analysis by the computer, as a rating is not possible. This knowledge will be carried onto test phase two, as the requirement for pictures that are to be used, is that they can all be analysed by the computer.

Directory: projekter -> files
files -> IT’d be weird without mcdonald’S
files -> Of Maj 2014 aau introduction & Theory
files -> Representations of Gender in Agatha Christie’s Poirot and Miss Marple
files -> Research in contemporary social movements: a case study of Guatemala 2015
files -> Anne Hviid-Pilgaard Master Thesis 31/05 2012 Table of Contents
files -> Jakob Nors-Ganer Aalborg University 28-05-2015
files -> Abstract: Purpose
files -> Security and trust in IoT/M2m cloud based platform
files -> The impact of french and british colonial rule on the he system in cameroon from the perspective of students
files -> Rumble in the jungle the ‘Blessing’ and ‘Curse’ of Mineral Wealth in the Congo

Download 0.97 Mb.

Share with your friends: