Humans have a remarkable capacity to learn and use facial motion to extract personal characteristic to recognize another individual’s facial motion which is considered is one of the most active areas of research in the field of computer vision and pattern recognition and numerous algorithms have been proposed to handle various aspects of the problem such as illumination, pose, expression, age (Aggarwal, Biswas , Flynn, & Bowyer, 2011), and even smile and human identity (Ning & Sim, 2010; Avargues-Weber , 2012).Facial motion is able to provide identity-specific dynamic facial signatures that benefits the perception of identity (Roark, et. al, 2003). For example, study has shown that people who are famous, their faces were able to be recognizes by the participant better when it is a moving image which involving several kind of facial motion rather than static images (Ning & Sim, 2010). Similar observation has been made by Thornton & Kourtzi (2002) too. Pilz, et al. also further explains that moving images not only help human subjects to improve their rate of recognition but also time of reaction too (Ning & Sim, 2010). This clearly shows that faces are a special type of visual pattern for which we haveunique expertise (Avargues-Weber , 2012). These findings may means that facial motion may contain considerable identity-specific information and that humans are probably using these to recognize each other in daily life (Ning & Sim, 2010).
In short, when increased response times or error rates of faces are shown in an inverted, but not upright orientation, the recognition performance of an individual will be impaired as it required strenuous amount of effort in recognition, this occurrence is known as the face inversion effect (FIE) (Martin & Macrae, 2010). This “face-inversion effect” (FIE) has become one of the standard tools for exploring face processing, in particular the roles of configure or relational versus feature (Thornton, et. al, 2011). Macrae and lewis have shown evidence that face recognition is caused by individuals’ orientation to global or due to local facial features during encoding of stimulus (Martin & Macrae, 2010). To add on, individuals’ propensity are different in order to identify complex visual stimuli in a global or local manner (Martin & Macrae, 2010).
Looking at the research question, there are actually a total of four hypotheses, to show a further understanding. The first, it is hypothesized that the mean scores of low static upright orientation are equal to that of the mean scores of the low static inverted orientation. Second, the mean scores of the low dynamic upright orientation are lower than the mean scores of the low dynamic inverted orientation. Third, the mean scores of the high static upright orientation are higher than the mean scores of the high static inverted orientation. The four and the last is that the mean scores of the high dynamic upright orientation are higher than the mean scores of the high dynamic inverted orientation.
150 participants from a University, including 39 males and 111 females. The students’ ages ranging from 19 to 51 with the mean age of 22.2. Students were supposed to complete the experiment, as part of an academic requirement of a tutorial class. All participants given written consent prior before commencing the research study.
In order to measure the level of facial motion in affecting face inversion effect, a Match- to- Sample task was used. Participants were required to fill up and complete the sequence of trials, identifying the right faces which was presented to them at the beginning of the experiment, followed by a blank screen of 5 seconds. Next step to do, they were to identify the right upright or inverted static and dynamic faces shown in each trial. The images of the faces will stay on screen till a response was given.
Participants were instructed in the beginning of the experiment how trial sequences will be presented. Four different sequence of trial types was randomly assigned in the presentation. The first trial type is an upright static image of the target face was presented to the participants. Second trial type, an inverted static image of the target face. Followed by the third trial type, an upright dynamic footage of the target face and the fourth trial type; an inverted dynamic footage of the target face. In each and every trial sequence, participants were presented with the image for 5 seconds, then followed by a blank screen for 5 seconds. Participants, next, were shown with two test images choices and there is only one of them matched the actual shown earlier which is correct. Images in the test will appear until the point of a response is being clicked on. The next trial test will then began with another new image right after a response was made.
The independent variables in the research study were Motion, being it dynamic or Static, Face Recognition skills either High or low recognition of faces and Orientation of the face, be it Upright or Inverted. The dependent variable in the research study will be the proportion of matches of the images that were deemed correct.
The results on the test were calculated based on accuracy of data collected and the performance to respond to test images and footage in an upright static condition using a one- way repeated measures analysis of variance test. These results were according from the top and bottom, participants’ result was evaluated based on the performance of the participants on the test in the static upright environment.
The mean proportion of correct matches in the low group participants who has poor facial recognition (FR) skills in the upright static environment (M= 0.82, SD=0.052) and inverted environment (M= 0.78, SD=0.134) was compared to high group participants who have high FR skills in the upright static environment (M= 1.0, SD=0.00) and inverted environment (M= 0.89, SD=0.096).
Figure 1. Mean proportion of correct matches made in the low and high groups under Static orientation
The result shows that the low group participants reflected poor FR skills in the upright static environment (M = 21.6 years, 6 males), t (48) = 16.88, p<.001, with an average mean score of .82. Where participants in the High group with the similar conditions (M= 24.7, 5 males) scored an average of 1.0. Figure 1 above summarises the mean proportion of correct matches of the low versus high facial recognition skills and upright versus inverted level of orientation. As Figure 1 has shown, the mean proportion of correct matches made in the low and high groups under Static orientation where the low FR for both upright and inverted orientation were not equal, this contradicted the first hypothesis where the mean scores of low static upright orientation to be equal to that of the mean scores of the low static inverted orientation.
Paired Sample Statistics for Low and High Facial recognition in upright and inverted orientations with static or dynamic conditions.
Table 1 above has shown that the mean proportion of correct matches made in the low and high facial recognition groups with static or dynamic condition. Based on the first hypothesis, it is not true as the results has shown in Table 1 that it does not support the overall studies done on the facial recognition and visual orientation. However, The mean scores of the high static upright orientation were higher than the mean scores of the high static inverted orientation which resulted that the third hypothesis which was mentioned earlier in the introduction is true.
Figure 2. Mean proportion of correct matches made in the low and high groups under Dynamic orientation.
Figure 2 above has shown that the mean proportion of correct matches made by low and high groups participants under Dynamic orientation. Based on the second hypothesis, it stated that mean scores of the low dynamic upright orientation were lower than the mean scores of the low dynamic inverted orientation. However, it seems that the graph in Figure 2 has shown differently where actually the mean scores of the low dynamic upright orientation was higher than the mean scores of the low dynamic inverted orientation. Therefore, it seems that second hypothesis which is based on the results shown on table 1 above is also not true and subsequently does not support the overall studies done on both facial recognition and visual orientation.
The means proportion of the graphs in the high static upright orientations is higher than the mean proportion of the graphs in the high static inverted orientation which is also reflected in Table 1. In the results of the final hypothesis as shown in Figure 2, the mean scores in the high dynamic upright orientation appeared to be higher than the mean scores of the high dynamic inverted orientation. This further concludes that the third and fourth hypothesis is true as reflected in Figure 1, 2 and Table 1 and supports the overall studies on facial recognition and visual orientation.
Dynamic Upright- Dynamic Inverted
Dynamic Upright- Dynamic Inverted
Table 2. 2 Paired Sample t test for Low and High Facial recognition in upright and inverted orientations with static or dynamic conditions.
Above Table 2 shows the results of mean proportion of correct matches made in the low and high groups of the participanta under the several types of visual orientation.
The aim of the study was to see how motion improved performance for people who have poor face recognition skills. The Participants were asked to complete the tests which involved series of trials that consist of differentiating the images of faces in upright and inverted orientations under static and dynamic conditions. Four hypotheses were created in conjunction to the research question given, to show further understanding of research question. The first hypothesis is that motion and recognition are proportional in terms of identifying faces, the dynamic movements able to help the people with poor recognition skills to identify motion. This meant that the upright condition is similar to the inverted condition, where both were static and the condition of low FIE, prove that it is not true. The second hypothesis where the mean scores of low dynamic upright orientation is much higher than the low dynamic inverted orientation in recognizing the faces in the trials, but the low dynamic inverted orientation group, the mean scores are higher. Therefore, the second hypothesis was proved to be not true, in terms of the low dynamic inverted orientation. Thus, it helps those poor recognition skills participants to perform better in an inverted orientation instead.
Using the result, the third and the last hypothesis shows that both upright orientations with static and dynamic conditions are greater than the inverted conditions which shows that facial recognition is assisted by the images in the upright orientation. The last hypothesis which the mean scores of the high dynamic upright orientation is proved to be true to be true that the scores were higher than the means scores in high dynamic inverted orientation, based on Longmore & Tree, 2013, this was demonstrated to be a benefit in aiding the process of face processing and facial recognition. Thus this concludes that motion does have a significant effect in the facial recognition process of people.
On contrary to the study done in discussion of the strengths and limitations of this study, it is found that the number of participants were a total of 150 which satisfy the general requirement of a study in having at least 30 participants. Therefore the criteria on sample size were fulfilled. As this study was done based on the focus that if motion was helpful, in terms of the targeting the right audience, this study was helpful in terms of helping individuals understand the issues with facial recognition and how the difference in orientation would affect our visual perception.
The limitations of the study is that it was conducted in the way for individual to understand that how motion actually helps in individuals’ recognition skills, involving studies of individual who have different cognitive perceptual disorders. Therefore, making the study seem unhelpful in giving information on how can motion be fully utilized in terms of treatment of the disorders. Another limitation would be the age ranges of participants, where it is between 19 to 51 ranges of age. It was proven that younger observers were proven to be better than older observers in multiple motion. Older observers usually require practice to be better due to controls of optical blur and retinal image even though they show similar learning function as the younger ones. (Legault, Allard, & Faubert, 2013).
To conclude, the result shows that motion do have an impact to in visual processing and facial recognition. However, Future studies should be conducted in a way where by the age range is conducted along with a balanced number of participants in term of gender. It would be better if future studies were to focus on specific audiences with more research and academic information as to understand better of the issues treating disorders. Last but not least, there is a level of significance showed in the student, proving that there are important information provided in regards to facial recognition skills.