Face Recognition - Essay Example

FACE RECOGNITION Face recognition is chiefly used in forensic science. The technicalities in recognition process are the ongoing research modalities. The use of face recognition in judiciary involves both the identification part of the process and the confidence part of the eye witness. The relation between the accuracy and confidence are being subjected to many research. Researchers have commonly accepted the fact that accuracy and confidence are loosely tied with one another. The purpose of this paper is to identify any more factors to be included in the optimality hypothesis so that the hypothesis can be put in to further tests. Thirty two students viewed face stimuli consisting 36 photographs during study phase and 76 photographs in test phase. They were asked questions about the general characteristics and distinctive features of the faces they viewed. Their YES/NO answers were found to confirm the fact that accuracy had a strong and positive relationship with confidence whereas confidence was not found feasible to determine the accuracy. FACE RECOGNITION INTRODUCTION: Face recognition is closely related to the accuracy of identification. Researchers are working on the subject of the confidence of eye witness and its relationship with the accuracy. Assessment of accuracy of eye witness by the jurors on the basis of the confidence both expressed and implied by the eye witness naturally depends upon the jurors angle of approach and direction of focus. Empirical evidence indicates that the relationship between an identification and a persons confidence in that direction is quite weak(Bothwell, Deffenbacher & Brigham, 1987; Wells and Murray, 1984). The optimality hypothesis proposed by Deffenbacher postulates that the relationship between the accuracy and confidence be the function of optimality of the conditions under which the faces are to be observed, identified and finally confirmed with confidence. The part where the confidence of the observer is analyzed encompasses several factors like the observer’s clarity in thoughts, absence of any external pressure to express the identification, presence of other variables like sound and movement and above all the time lag between the test phase and study phase. The findings of Winograd (1981) revealed that face recognition is improved , if during the study phase, the participants are required to answer questions about the distinctiveness of the face or specific traits, compared to the questions about a single physical feature. During the recognition test participants were asked to decide whether a test face had appeared during their study phase. And to indicate their confidence level. Their YES/NO answer for identifying a test phase is followed by the question of expressing their confidence level. The level of confidence derived by their replies are to be judged in the light of their YES/NO reply. This type of judging the confidence level leads to less strength to the findings. The desirability of using more than one measure of eye witness confidence in attempting to predict accuracy leads to test and retest procedure and Interjudge reliability method. The test – retest method yielded a an average correlation of 0.91 between two administered measures. The inter judge relying method, wherein the extent of agreements of more than two individuals who independently assign scores to the target cases are measured, although leads to reliable scores, does not provide any useful discrimination. This is because the agreement does not imply that the confidence of the observer does not reflect their accuracy. METHOD: Participants: Fifty participants were attending the study and test phase of face recognition process. Of these 14 were males and 36. Design: The independent variable was questions asked during encoding. This factor was spread among the participants by asking them to delineate the physical feature or distinctiveness of the face they observed during study phase. The dependant variable was the recognition performance which was got elucidated by their YES/NO replies to the question if they happened to see any face in the test phase, which were already beheld by them in the study phase. Materials: the materials used for the test are the face stimuli consisting of photographs of Caucasian female faces of bust size. The age group of the faces were selected to fall between 18-20. Care was taken to see that no easily identifiable features are found in the images, like glasses, hairstyles and jewellery. 36 images of female faces described above were used for study phase and 72 female face images including the study phase images were used for test phase . The purpose of including new or distracter faces was to test the confidence rating of the participants. Experimental booklets including a cover sheet, study questions response sheets and scoring protocol were disbursed among the participants. They were given a set of study questions; one question for each face, numbered 1-36. the questions were randomized. Three examples of each question was repeated 6 times in the following manner. Physical feature condition: - Does this face have a big nose? Does this face have close-set eyes? Does this face have thin lips? Distinctive feature condition: - Is the nose the most distinctive feature? Are the eyes the most distinctive feature? Is the mouth the most distinctive feature? The response sheets consisted of item numbers 1 to 72 with the word YES and NO and the confidence scale printed beside each item number. A 5 point Likert scale was used in the confidence ratings. (1 = not at all confident & 5 = totally confident). The scoring protocols replicated the response sheets and were coded in such a manner that each number identified with appropriate condition like old or new face, physical or distinctive feature condition. A similar set of study and test was conducted and the findings compiled together under version A and version B in order to have control for any variation that may result from a particular face being more or less easily recognized. Procedure: The participants were allowed to view for 5 seconds the photographs during their study phase and for 10 seconds during the test phase. During the study phase the participants were told that they were participating in a memory experiment, and that they are required to answer one question for each face as it is presented. The recognition test required participants to respond YES if they think the face appeared during the study phase, and NO if they do not think the face occurred before, and to indicate their confidence in the decision using the rating scale provided. RESULTS AND DISCUSSION: The purpose of this paper is to find out any factor that can be included in optimality hypothesis. On going through the literature and the results it is found that the optimality factors are minimal in arriving at the judgment on the basis of confidence. The confidence factor expressed by the observer is subjected to variables that often deviate from their confidence and sometimes dilute the confidence. For example, when an observer is asked a repeated question to get a clear confirmation of his/her confidence shakes the confidence itself and the observer produces less confident feedback. Deffenbacher’s optimality hypothesis still remains un questioned as it generalizes that the relationship between accuracy and confidence be the function of optimality, the ‘condition’ of which the faces are to be observed, identified and recognised. Winograd (1981) and Zajonc (1980) findings are seemingly contradicting each other. While Zajonc emphasis that the type rather than the amount of encoding that accounts for recognition of faces, Winograd states that it is the amount of encoding rather than the type of encoding that matters in judging the accuracy of face recognition. When these two statements are analyzed minutely the externally seeming contradiction disappears. Factually both the findings are inter related. The more the amount of encoding is used, the process as a whole implies certain characters/traits of the faces thus assuming a fresh dimension in the type of encoding. Likewise, the type of encoding while accounting for judgments, simultaneously carries along with it the information/feature/trait in a capsule form thus providing them a status of constituents to the process of encoding. The findings of the study are tabulated as below; DATE AGE GENDER POH:PF POH:DF FALSE ALARMS MCR:Correct PF MCR:Incorrect PF MCR:Correct DF MCR:Incorrect DF Thu Sep 12 11:56:03 2002 22yrs:2mths F 0.67 0.56 0.31 4.4 2.75 3.67 3.5 Thu Sep 12 12:13:47 2002 18yrs:8mths F 0.61 0.83 0.03 4.73 4.33 4.18 4.14 Thu Sep 12 12:46:55 2002 18yrs:4mths F 0.83 0.83 0 4.07 2.67 3.67 3.33 Thu Sep 12 12:58:50 2002 19yrs:2mths M 0.61 0.83 0.14 4.4 3.67 3.73 3.43 Thu Sep 12 13:17:26 2002 19yrs:1mths M 0.89 0.78 0.22 4.43 4.25 4.56 4.5 Thu Sep 12 13:28:06 2002 18yrs:1mths F 0.83 0.83 0.42 4.6 2.33 4.33 2.67 Thu Sep 12 13:28:13 2002 18yrs:5mths M 0.39 0.56 0.03 4.8 3.88 4.57 4 Thu Sep 12 13:29:47 2002 33yrs:1mths M 0.67 0.72 0.19 3.38 3 3 2.33 Thu Sep 12 13:33:58 2002 18yrs:6mths F 0.67 0.72 0.44 4.54 3.6 3.67 3.83 Thu Sep 12 13:34:00 2002 19yrs:9mths M 0.44 0.67 0.14 3.58 2.5 3.25 3.3 Thu Sep 12 13:42:59 2002 20yrs:5mths M 0.56 0.72 0.22 4.77 3.8 4.9 4 Thu Sep 12 13:50:50 2002 18yrs:5mths F 0.44 0.72 0.36 4.46 2.6 4.13 3.3 Thu Sep 12 13:52:50 2002 19yrs:6mths F 0.61 0.44 0.44 5 4.7 5 4.86 Thu Sep 12 13:53:39 2002 19yrs:5mths F 0.61 0.72 0.08 4.38 3.8 4.09 3.71 Thu Sep 12 13:58:20 2002 20yrs:4mths F 0.56 0.61 0.31 2.27 1.43 2.9 1.88 Thu Sep 12 14:26:45 2002 19yrs:3mths F 0.56 0.78 0.11 4.29 3.25 3.9 3.63 Thu Sep 12 14:38:52 2002 19yrs:4mths F 0.67 0.78 0.06 3.86 3.75 4 3.17 Thu Sep 12 14:53:02 2002 21yrs:1mths F 0.61 0.89 0.08 4.81 4 4.73 3.71 Thu Sep 12 15:15:32 2002 22yrs:3mths F 0.39 0.89 0.03 4.25 2 4.14 4.27 Thu Sep 12 15:26:53 2002 19yrs:7mths F 0.67 0.89 0.22 4.19 3 4 3.67 Thu Sep 12 15:35:13 2002 21yrs:2mths M 0.72 0.39 0.31 4 3.55 4.31 3.2 Thu Sep 12 15:38:27 2002 19yrs:7mths F 0.78 0.83 0.25 3.6 3.33 3.43 3 Thu Sep 12 17:27:32 2002 19yrs:5mths F 0.44 0.44 0.19 5 5 5 5 Fri Sep 13 08:13:52 2002 18yrs:5mths F 0.83 0.83 0.08 3.4 2.33 3.87 2 Fri Sep 13 10:18:14 2002 19yrs:5mths F 0.44 0.67 0.03 4.08 2.67 3.63 2.9 Fri Sep 13 11:19:21 2002 21yrs:1mths F 0.33 0.67 0.14 2.92 1.83 3.33 1.5 Fri Sep 13 11:33:00 2002 18yrs:6mths F 0.61 0.83 0.22 4.87 5 5 5 Fri Sep 13 11:53:07 2002 19yrs:6mths F 0.83 0.67 0.39 4.75 3.67 4.4 3.67 Fri Sep 13 12:11:13 2002 30yrs:4mths F 0.28 0.44 0.03 4.75 4.9 4.8 5 Fri Sep 13 12:34:20 2002 20yrs:3mths F 0.5 0.83 0.03 4.6 2.67 4.22 3.33 Fri Sep 13 12:45:31 2002 20yrs:9mths F 0.61 0.83 0.03 4.93 5 4.27 3.86 Fri Sep 13 13:02:27 2002 18yrs:2mths M 0.56 0.5 0.53 3.33 3.44 3.3 2.63 Fri Sep 13 13:07:51 2002 18yrs:2mths M 0.44 0.5 0.53 3.22 3.11 3 2.9 Fri Sep 13 13:10:33 2002 19yrs:2mths M 0.61 0.89 0.11 4.69 4 4.36 4.43 Fri Sep 13 14:54:52 2002 18yrs:5mths F 0.61 0.67 0.11 4.67 3 4.64 3 Fri Sep 13 14:54:55 2002 18yrs:4mths F 0.56 0.61 0.22 4.82 4 5 3.38 Fri Sep 13 15:01:22 2002 19yrs:4mths F 0.61 0.94 0.17 4.41 2 4.36 3.86 Fri Sep 13 15:24:47 2002 19yrs:5mths M 0.89 0.83 0.11 5 3.33 5 4.5 Sun Sep 15 15:47:56 2002 yrs:8mths F 0.89 0.94 0.08 4.47 3 4.56 4 Mon Sep 16 09:29:22 2002 18yrs:3mths F 0.78 0.72 0.19 3.54 1.4 4.14 3.25 Mon Sep 16 10:26:10 2002 20yrs:3mths F 0.44 0.78 0.03 4 2.75 2.88 2.9 Mon Sep 16 15:11:52 2002 18yrs:3mths F 0.44 0.72 0 4 2.8 4.25 3.7 Mon Sep 16 16:17:19 2002 19yrs:2mths M 0.72 0.94 0.17 3.65 2 3.23 2.8 Tue Sep 17 09:29:29 2002 18yrs:9mths F 0.67 0.61 0.19 4.64 3.71 4.75 3.67 Tue Sep 17 14:27:30 2002 18yrs:9mths F 0.83 0.94 0.22 4.71 3 4.6 3 Tue Sep 17 15:32:19 2002 22yrs:8mths F 0.94 0.72 0.14 4.62 3.4 4.35 2 Tue Sep 24 13:09:39 2002 19yrs:1mths F 0.28 0.44 0.22 4.5 3.8 4.2 4.15 Tue Sep 24 13:50:23 2002 20yrs:9mths F 0.83 0.78 0.14 4 3 3.33 1 Wed Sep 25 09:30:08 2002 31yrs:4mths M 0.5 0.94 0.08 4.82 4 4.89 3.89 Wed Sep 25 13:25:01 2002 20yrs:1mths M 0.17 0.39 0 4.86 2.45 2.67 2 As regards the trait, so many factors come in the way before reaching a confident judgment. The factors in relation to physical features of the faces may be fixed and taken into account for identification. But the factors that characterize the trait of the observer and the observed could not be limited to any lab atmosphere. Especially the time factor plays a crucial role. The confidence rating in a test phase is liable to be certainly affected when the time gap between these two phases is lengthened. To arrive at a clear cut finding, a separate study of the same can be carried out interchanging the time factors of 5 seconds and 10 seconds for test phase and study phase. Similarly the movement and voice factors of the faces to be identified occupy a crucial place in the process of recognition underlying key functions of the traits. Recollecting an image from a static photograph is relatively less probable than recollecting a face that winkles in a motion picture. The voice factor definitely affects the judgement in the level of processing. Subtle changes in decibel are likely to produce negative results. Thus inclusion of factors like time lag and movement facilitate further venue for research, whereas the sound factor reduces the option for arriving at a candid source of findings. In conclusion, it is clearly found that only a nearness to accuracy can be obtained and not the accuracy of cent per cent perfection can be reached. Reference list – Bothwell, R.K. Deffenbacher, K.A. & Brigham J.C. (1987) “Correlations of eyewitness accuracy and confidence: Optimality hypothesis revisited, Journal of Applied Psychology, 72,691. Wells, G.L. & Murray, D.M. (1984). Eyewitness confidence. In G.L.Wells & E.F.Loftus(Eds), Eyewitness testimony: Psychological perspectives(pp 155-170), New York: Cambridge University Press. Winograd, E. (1981), Elaboration and distinctiveness in memory for faces, Journal of experimental Psychology: Human Learning and Memory, 3, 181-190. Zajonc, R.B. “ Feeling and thinking”, Preferences need no Inferences, American Psychologist, 1980, 35, 151-175. Read More