Samsung have developed a smart phone that will identify from such various indications such as writing speed or weather conditions, the emotions of a user.
The impassive computer, generations of users with cold logic and obscure error messages irritated has, could soon be a thing of the past. Researchers at the ElectronicConcerns Samsung have developed a smart phone that will recognize the emotional state of a user - not by means of sensors or cameras, but by the way the phone is used. serve as a guide as data on how quickly a user character enters, as he often the "Reset" - or another special press buttons, and how strong the unit is moved back and onward. From software calculate whether the phone carrier was happy, angry, cheerful, or surprised, says Hosub Lee from Samsung Advanced Institute of Technology in South Korea, who led the research project.
Also give the impression such measures do at first glimpse have little to emotions, there are slight connection between them and the emotional state of a user. The software algorithms use the so-called machine learning, and currently have a hit rate of 67.5 percent, Lee said. Their prototype, the next week at the Consumer Communications and Networking Conference will be presented in Las Vegas is part of a Twitter application for The Samsung Galaxy S II This allows users of social networks complement their text messages to icons that show emotions. There are still plenty of other applications are conceivable, says Lee. Here is what the emotion recognition choose different ringtones that fit the particular mood. "The smart phone users could also play a bad mood in a cartoon to cheer him," says Lee.
But that was just the beginning. "Emotion recognition is the door to sophisticated equipment or services open to recognize the contexts", Samsung expects the researcher. "If we know the current emotional state of a user, we can offer him the services that are bespoke specifically to him." The software must be trained first in order to analyze the characteristics of a user. In this phase it is characterized at each tweet to a user settles on a variety of sizes, could be the indication of a certain emotion. This includes the current weather or light conditions. The user records his part for each short message, how he feels.
All this data is fed into a so-called Bayesian network. Where it is identified is a machine learning algorithm, the correlations between apparently unrelated data. In a Bayesian network represent measured variables represent nodes that can be connected to each other. Each value of a measure - for example, the fast and the slow typing on the display is - is assigned a probability that it further, with their associated metrics, influenced. From the interplay of several popular sizes is then mathematically a probability that the requested variable - in this case, a mood as "excited" - is present
The technology was still at an early experimental phase. So far, only a single user inputs were tested. Once the researchers want to collect more training data."So that we can discover new clues for certain moods," says Lee. He hopes that the system can make some assumptions about personality, even when a user. Rosalind Picard, one of the leading researchers in the field of emotion recognition, sees the process of the Samsung Group; a new approach should continue to follow the man. "There is a great demand for technologies that can enable organizations to enter into the feelings of their customers," the director of the Affective Computing Research Group at MIT. "To recognize when a customer is interested or bored, confused or pleased, marks an important first step in to treat him with respect."