Requirement Specification

6.2.Programming Requirements

The software would have to output a numerical value that corresponds to the smile rating. The numerical rating can then be referenced against the answers given by test subjects to understand what differentiates the two. As this is an early attempt of smile understanding and interpretation by a computer, the speed or efficiency of the algorithm is of lesser concern, as the program is to be operated by the author of this thesis only.

7.Implementation

The implementation follows the structure of the requirement specification with the addition of the analysis on the pictures selected for test phase one.

Pictures used in Phase One based on the picture test:

Figure - Test Phase One – Pictures, in order of appearance, top left (Picture 1) to bottom right (Picture 26)

The selected pictures from Figure 1 were taken from Google image, Flickr and other photo sharing sites. Each picture was selected from a batch of +50 pictures. The requirement for each picture selected was that the expression portrayed by the subjects in the pictures had to be in the range of either completely happy or completely sad. By having diametric opposites of expressions, the expressions could then be converted to a numerical scale of 1-10. Since the author of this thesis selected the pictures in the initial batch (50 pictures), they had to be given to a test group to establish that the emotions portrayed in the pictures were real and indicative to the happy to sad range. Therefore each picture was given a dry test: each picture was printed in colour and given to a preliminary test group. Each test subject was asked to rate the level of happiness in the picture; this was to establish that bias from the author, if any, would be removed, and to insure that the test group could relate emotionally to the pictures, else a rating would not be possible. Of the 50 pictures, 20 pictures of the batch where chosen by the test group (depicted in Figure 1). The pictures in figure 1 appear in order of appearance in the test, top left is the first picture presented to the test group, bottom right last picture presented to the test group.

Figure - Welcome screen, Gender - Test Website

The website created for the test was written in Action Script 3 (AS3), for complete source code see Appendix 16.7. Action script 3 was chosen due to the author’s affinity with Action Script and AS3’s fairly easy implementation in regards to PHP and xml. Test participants, upon entering the URL of the website, were greeted with a welcome screen shortly explaining the task at hand. Test participants were required to enter their gender followed by submitting their age. Each input was written to an xml file residing on the webserver, the following code snippet is an answer given by a test participant taken directly from the xml file: “m 41 6 4 5 7 10 3 10 10 8 10 2 1 7 2 6 9 8 7 3 9 4 10 5 3 8 4”. The “m” denotes male and the first number the age. Each subsequent number i.e. 6 – 4 – 5 – 7, is the rating given by the test subject to the first, second, third and fourth picture ending with a rating of 4 for the last picture. In the introduction it was stressed that no right or wrong rating exists, since the goal of the test is subjective and each rating given is the absolute truth for the particular test participant.