HCI
Wexelblat' questions this point of view and reports on a preliminary study that brings the
anti-anthropomorphic argument into question. The experiment involved users performing tasks presented to them with different interface. In genrel, the deal of (sometimes heated) discussion in recent years among interfaces designers and researchers. (As Wexellblat writes,"Don't anthropomorphize computers; they hate that!")
This debate may be somewhat of a red herring. When a computer is seen as a tool-eg, a device used to produce a spreadsheet for data analysis - the anti-anthropomorphic argument is convincing. Users would not want a humanoid spreadsheet interface to be unpredictable when entering values or calculating sums, for example. or when moving cells to a different column. However, when computers are viewed as media or collaborators rather than as tools anthropomorphic qualities may be quite appropriate. Tool and tasks that are expected to be predictable should be so - but as we move away from office productivity applications to more pervasive use of computers, it may well be that the requirements of predictability and direct manipulation are too limiting
Concerning the second question, which are the most promising technologies, there are several examples of promising work and prototype systems that may help give indications. For example, the QuickSet system at OGI is an architecture for multimodal integration, and is used successfully for inte grating speech and(pen) gesture as users create and control military simulations. Another system for integrating speech and (visual) gesture is described by Poddar, et al. , applied to parsing video of a weather report. Another example of tight integration between modalities is in the budding "speechreading" community. These systems attempt to use both visual and auditory information to understand human speech - which is also what people do, especially in noisy environments.
4.Vision. Based Interaction
Present-day computers are essentially deaf, dumb, and blind. Several people have pointed out that the bathrooms in most airports are smarter than any computer one can buy, since the bathroom"knows" when a person is using the sink or toilet. Computers, on the other hand, tend to ask us questions when we're not there(and wait 16 hours for an answer) and decide to do irrelevant(but CPU- intensive) work when we're frantically working on an overdue document.
Vision is clearly an important element of human- communication. Although we can communicate human without it, people still tend to spend endless hours travelling in order to meet face to face. Why? Because there is a richness of communication that cannot be matched using only voice or text. Body language such as facial expressions, silent nods and other gestures add personality, trust, and important information in human-to- human dialog. We expect it can do the same in human computer interaction.
Vision based interfaces(VBI) is a subficld of perceptive media which concentrates on developing visua awareness of people. VBI seeks to answer questions such as:
- Is anyone there?
- Where are they?
- Who are they?
- What are the subject's movements?
- What are his facial expressions?
- Are his lips moving?
- What gestures is he making??
These questions can be answered by implementi computer vision algorithms to locate and identify individuals, track human body motions, model the head and face, track facial features, interpret human motion and actions. (For a taxonomy and discussion of movement, action, and activity, see Bobick).
VBI t(and, in general, PUls) can be categorized into two aspects: control and awareness. Control is explicit communication to the system e.g, put that object there. Awareness, picking up information about the subject without an explicit attempt to communicate, gives context to an application(or to a PUI). The system may or may not change its behavior based on this information. For example, a system may decide to stop all unnecessary background processes when it sees me enter the room- not because of an explicit command I issues, but because of a change in its context. Current computer interfaces have little or no concept of awareness. While many research efforts emphasize VBI for control, it is likely that VBI for awareness will be more useful in the long run.
The remainder of this section describes VBI projects to quickly track a user's head and use this for both awareness and control(Section 4.1), recognize a set of gestures in order to control virtual instruments(Section 4.2), and track the subject's body using an articulated kinema