Tahia Infantes Morris

CS7322

          "Week 1 - Critique of Rosenfeld Paper on HAL"

        So, now that I have seen the movie, I understand a little more about HAL - the emotional system. Hal supposedly was such an advanced system that it was able to interpret human emotions and lip read. I think not. First of all, it is very difficult, as it is, for people to interpret other people's emotions based on images alone. More background information about the person is needed to decide whether a person is smiling and happy versus hysterical and upset. I really love the quote "I can SEE that you're really upset". Not only is the word "SEE" used, but on top of that, HAL qualifies the statement with "REALLY UPSET". I watched the movie and the truth is that Dave seemed rather expresionless the entire movie. - muchless "REALLY" upset. If HAL relied on vision to interpret emotion, then why did HAL deny Dave entrance back into the spacecraft when Dave "hid" his anger? My next point about lip reading is kind of obvious. I outlined that thought in my last critique the previous quarter, so I will skip the commentary. The next issue that concerns me (before I actually write about what was assigned) is "How did HAL know that the sketch was, in fact, a sketch versus a photograph of Dr. Hunter?" How did HAL know that the rendering was an improvement?

        Okay, now on to the real question: "what does it mean to recognize expressions using image motion?" Well, the paper outlines how in detecting distortions in specific facial features (ie. the mouth, eyes, eye brows), a vision system can reliably identify the expression. Basically, the hardest part is in identifying and tracking the features - especially when the subject is moving about. Once the features have been identified, then certain values associated with the features are measured from frame to frame - such as curvature of the mouth and eye brows. The measurements are then compared to some kind of known list or database to extract the "emotion" that closely fits. This in theory sounds straight forward, but that assumes that the person is not trying to hide his/her emotions. I still not believe that by using image information alone, a vision system can determine a person's intentions. That is ludicrous. I believe that other information is required like electrical impulses and voice intonation. (I think we should try placing electrodes in a person's head.. buahahahahaha)