Ross-Flanigan, Kognitywistyka
[ Pobierz całość w formacie PDF ]
//-->MUSCLEHEADSOur faces have 44 muscles, most of whichcontract in a single way. Some muscles,such as those in our foreheads, can movein several ways. Scientists have definedthese various movements in “facial actioncodes,” which they program computers torecognize. From Paul Ekman & WallaceFriesen,The Facial Action Coding System,originally published by ConsultingPsychology Press, 1978.TR U T HA New Tool to AnalyzeOur ExpressionsBy Nancy Ross-FlanigantheFacingAcurled lip, a furrowed brow—sometimes even a smallchange in expression can reveal far more than words.We all like to think we can read people’s faces for signs oftheir true emotions. Now, a computer program can analyzeimages of faces as accurately as trained professionals.What’s more, it does so faster. Working frame by frame,the most proficient human experts take an hour to code the1,800 frames contained in one minute of video images, a jobthat the computer program does in only five minutes. A teamled by HHMI investigator Terrence Sejnowski reported theHHMIBULLETINM AY200113feat in the March 1999 issue of the jour-nalPsychophysiology.The automated system, which has beenimproved since the article appeared, couldbe a boon for behavioral studies. Scientistshave already found ways, for example, todistinguish false facial expressions of emo-tion from genuine ones. In depressedindividuals, they’ve also discovered differ-ences between the facial signals of suicidaland nonsuicidal patients. Such researchrelies on a coding system developed in the1970s by Paul Ekman of the University ofCalifornia, San Francisco, a coauthor ofthePsychophysiologypaper. Ekman’sFacial Action Coding System (FACS)breaks down facial expressions into 46individual motions, or action units.Sejnowski’s team designed the computerprogram to use the same coding system.Their challenge was to enable the programto recognize the minute facial movementsupon which the coding system is based.Other researchers had come up with dif-ferent computerized approaches for ana-lyzing facial motion, but all had limita-tions, says Sejnowski, who is directorof the Computational NeurobiologyLaboratory at The Salk Institute forBiological Studies in La Jolla, California,and a professor of biology at theUniversity of California, San Diego(UCSD). A technique called feature-basedanalysis, for example, measures variablessuch as the degree of skin wrinkling at var-ious points on the face. “The trouble,”Sejnowski explains, “is that some peopledon’t wrinkle at all and some wrinkle a lot.It depends on age and a lot of other fac-tors, so it’s not always reliable.”His team—which included Ekman,Marian Stewart Bartlett of UCSD andJoseph Hager of Network InformationResearch Corp. in Salt Lake City—tookthe best parts of three existing facial-motion-analysis systems and combinedthem.“We discovered that although each ofthe methods was imperfect, when we com-bined them the hybrid method performedabout as well as the human expert, whichis at an accuracy of around 91 percent,”Sejnowski says. The computer programdid much better than human nonexperts,14HHMIM AYBULLETIN2001Sejnowski is most interested in usingthe system to explore informationprocessing in the human brain. He likesto think of the brain’s changing activitypatterns as “brain expressions,” similarin many ways to facial expressions.who performed with only 73.7 percentaccuracy after receiving less than an hourof practice in recognizing and codingaction units. The coding process involvesidentifying and marking sequences offrames in which an individual facialexpression begins, peaks and ends. Aminute of video can contain several hun-dred action units to recognize and code.In the work reported inPsychophysiology,the researchers taught the computer pro-gram to recognize 6 of the 46 actionunits. Since then, the program has mas-tered six more and, by incorporating newimage-analysis methods developed inSejnowski’s lab, the system’s performancehas risen to 95 percent accuracy. Theadditional work was published in theOctober 1999 issue ofIEEE Transactionson Pattern Analysis and MachineIntelligence.Now the team is engaged in a friendly“cooperative competition” with researchersfrom Carnegie Mellon University and theUniversity of Pittsburgh who have devel-oped a similar system. The two systems willbe tested on the same images to allow directcomparisons of performance on individualimages as well as overall accuracy. Theteams will then collaborate on a newsystem that incorporates the best features ofeach.A computer that accurately reads facialexpressions could result in a better lie detec-tor, which is why the CIA is funding thejoint project. But Sejnowski sees other pos-sible commercial applications as well.“This software could very well end upbeing part of everybody’s computer,” hesays. “One of the goals of computer sci-ence is to have computers interact withus in the same way we interact with otherhuman beings. We’re beginning to seeprograms that can recognize speech.”But humans use more than speech recog-nition when they communicate with eachother, he explains. In face-to-face conver-sation, “you watch how a person reactsto know whether they’ve understoodwhat you’ve said and how they feelabout it.” Your desktop computer can’tdo that, so it doesn’t know when it hascorrectly interpreted your words or whenit has bungled the meaning. With thissoftware and a video camera mounted onyour monitor, Sejnowski thinks yourcomputer might someday read you aswell as your best friend does.Sejnowski and his team are engaged in afriendly “cooperative competition” withother researchers, with whom they willcollaborate on improved systems.MARK HARMELAnalyzing Our ExpressionsBRIGHT LIGHTS, MOVING HEADSFAKE EXPRESSIONSA computer system has to be taught to “see” someone’s furrowing brow or crinkly smile.Sejnowski and his team have drawn from diverse fields of science, such as artificialintelligence and neural networks, to handle this task. Among the many hurdles they face ishead movements and varying light conditions. To overcome this, their system breaks downthe images of faces into tiny units, measures the light in each dot, and then compares themosaic to known facial patterns. It’s a process that resembles—but is far more primitivethan—the way our brains make sense of light passing through the eye. In these photos,the smile on the left is false while the smile on the right is real.BE VERY AFRAIDIf the outer part of your eyebrow goes up,you’re probably afraid—although you mayjust be feeling surprise. To determinewhether someone is truly afraid, like theman in this photo, Sejnowski’s team looksfor several other facial actions that indicatefear. All of the following movements in theupper part of the face may indicate fear,although some are also associated withother emotions:•Raising the upper eyelid to show thewhites of the eyes•Raising the inner brow:Also mayshow sadness•Lowering the brow:Also may showanger or mental effort•Tightening the eyelid:Also may showanger or disgustSeveral facial cues suggest that someone isfaking an expression, among them:• The muscle contractions on the left side ofthe face differ from those on the right side.• The expressions start and stop in a jerky manner.• The person holds the expression for too long.Eye movements also provide information.Someone looking downward may be sad, whilesomeone looking down or away is more likely feel-ing shame, guilt or disgust. The sad look on thisman’s face is false.16HHMIM AYBULLETIN2001 [ Pobierz całość w formacie PDF ]
zanotowane.pl doc.pisz.pl pdf.pisz.pl pingus1.htw.pl
//-->MUSCLEHEADSOur faces have 44 muscles, most of whichcontract in a single way. Some muscles,such as those in our foreheads, can movein several ways. Scientists have definedthese various movements in “facial actioncodes,” which they program computers torecognize. From Paul Ekman & WallaceFriesen,The Facial Action Coding System,originally published by ConsultingPsychology Press, 1978.TR U T HA New Tool to AnalyzeOur ExpressionsBy Nancy Ross-FlanigantheFacingAcurled lip, a furrowed brow—sometimes even a smallchange in expression can reveal far more than words.We all like to think we can read people’s faces for signs oftheir true emotions. Now, a computer program can analyzeimages of faces as accurately as trained professionals.What’s more, it does so faster. Working frame by frame,the most proficient human experts take an hour to code the1,800 frames contained in one minute of video images, a jobthat the computer program does in only five minutes. A teamled by HHMI investigator Terrence Sejnowski reported theHHMIBULLETINM AY200113feat in the March 1999 issue of the jour-nalPsychophysiology.The automated system, which has beenimproved since the article appeared, couldbe a boon for behavioral studies. Scientistshave already found ways, for example, todistinguish false facial expressions of emo-tion from genuine ones. In depressedindividuals, they’ve also discovered differ-ences between the facial signals of suicidaland nonsuicidal patients. Such researchrelies on a coding system developed in the1970s by Paul Ekman of the University ofCalifornia, San Francisco, a coauthor ofthePsychophysiologypaper. Ekman’sFacial Action Coding System (FACS)breaks down facial expressions into 46individual motions, or action units.Sejnowski’s team designed the computerprogram to use the same coding system.Their challenge was to enable the programto recognize the minute facial movementsupon which the coding system is based.Other researchers had come up with dif-ferent computerized approaches for ana-lyzing facial motion, but all had limita-tions, says Sejnowski, who is directorof the Computational NeurobiologyLaboratory at The Salk Institute forBiological Studies in La Jolla, California,and a professor of biology at theUniversity of California, San Diego(UCSD). A technique called feature-basedanalysis, for example, measures variablessuch as the degree of skin wrinkling at var-ious points on the face. “The trouble,”Sejnowski explains, “is that some peopledon’t wrinkle at all and some wrinkle a lot.It depends on age and a lot of other fac-tors, so it’s not always reliable.”His team—which included Ekman,Marian Stewart Bartlett of UCSD andJoseph Hager of Network InformationResearch Corp. in Salt Lake City—tookthe best parts of three existing facial-motion-analysis systems and combinedthem.“We discovered that although each ofthe methods was imperfect, when we com-bined them the hybrid method performedabout as well as the human expert, whichis at an accuracy of around 91 percent,”Sejnowski says. The computer programdid much better than human nonexperts,14HHMIM AYBULLETIN2001Sejnowski is most interested in usingthe system to explore informationprocessing in the human brain. He likesto think of the brain’s changing activitypatterns as “brain expressions,” similarin many ways to facial expressions.who performed with only 73.7 percentaccuracy after receiving less than an hourof practice in recognizing and codingaction units. The coding process involvesidentifying and marking sequences offrames in which an individual facialexpression begins, peaks and ends. Aminute of video can contain several hun-dred action units to recognize and code.In the work reported inPsychophysiology,the researchers taught the computer pro-gram to recognize 6 of the 46 actionunits. Since then, the program has mas-tered six more and, by incorporating newimage-analysis methods developed inSejnowski’s lab, the system’s performancehas risen to 95 percent accuracy. Theadditional work was published in theOctober 1999 issue ofIEEE Transactionson Pattern Analysis and MachineIntelligence.Now the team is engaged in a friendly“cooperative competition” with researchersfrom Carnegie Mellon University and theUniversity of Pittsburgh who have devel-oped a similar system. The two systems willbe tested on the same images to allow directcomparisons of performance on individualimages as well as overall accuracy. Theteams will then collaborate on a newsystem that incorporates the best features ofeach.A computer that accurately reads facialexpressions could result in a better lie detec-tor, which is why the CIA is funding thejoint project. But Sejnowski sees other pos-sible commercial applications as well.“This software could very well end upbeing part of everybody’s computer,” hesays. “One of the goals of computer sci-ence is to have computers interact withus in the same way we interact with otherhuman beings. We’re beginning to seeprograms that can recognize speech.”But humans use more than speech recog-nition when they communicate with eachother, he explains. In face-to-face conver-sation, “you watch how a person reactsto know whether they’ve understoodwhat you’ve said and how they feelabout it.” Your desktop computer can’tdo that, so it doesn’t know when it hascorrectly interpreted your words or whenit has bungled the meaning. With thissoftware and a video camera mounted onyour monitor, Sejnowski thinks yourcomputer might someday read you aswell as your best friend does.Sejnowski and his team are engaged in afriendly “cooperative competition” withother researchers, with whom they willcollaborate on improved systems.MARK HARMELAnalyzing Our ExpressionsBRIGHT LIGHTS, MOVING HEADSFAKE EXPRESSIONSA computer system has to be taught to “see” someone’s furrowing brow or crinkly smile.Sejnowski and his team have drawn from diverse fields of science, such as artificialintelligence and neural networks, to handle this task. Among the many hurdles they face ishead movements and varying light conditions. To overcome this, their system breaks downthe images of faces into tiny units, measures the light in each dot, and then compares themosaic to known facial patterns. It’s a process that resembles—but is far more primitivethan—the way our brains make sense of light passing through the eye. In these photos,the smile on the left is false while the smile on the right is real.BE VERY AFRAIDIf the outer part of your eyebrow goes up,you’re probably afraid—although you mayjust be feeling surprise. To determinewhether someone is truly afraid, like theman in this photo, Sejnowski’s team looksfor several other facial actions that indicatefear. All of the following movements in theupper part of the face may indicate fear,although some are also associated withother emotions:•Raising the upper eyelid to show thewhites of the eyes•Raising the inner brow:Also mayshow sadness•Lowering the brow:Also may showanger or mental effort•Tightening the eyelid:Also may showanger or disgustSeveral facial cues suggest that someone isfaking an expression, among them:• The muscle contractions on the left side ofthe face differ from those on the right side.• The expressions start and stop in a jerky manner.• The person holds the expression for too long.Eye movements also provide information.Someone looking downward may be sad, whilesomeone looking down or away is more likely feel-ing shame, guilt or disgust. The sad look on thisman’s face is false.16HHMIM AYBULLETIN2001 [ Pobierz całość w formacie PDF ]