TU-TESL

• 22

According to foreign media reports, Mohamed abouelenien, assistant professor of computer and information science at the University of Michigan at Dearborn in the United States, once entered the field of lie detection, created a prediction model based on face thermal imaging, and did similar research in the field of human alertness, and this technology is very popular in the field of automobile. A recent study by Professor Mohamed abouelenien has been linked to a particularly subjective feeling in human physiology: thermal comfort. When people are in a common space, such as home, office or car, it is difficult for people to reach an accurate consensus on what kind of temperature is more comfortable. And many people will find that a single temperature setting can not make everyone feel comfortable, which may lead to a tug of war on temperature regulators. It may be easier to customize in the field of automobile, and now many models provide passengers with methods to control their environment in the semi independent of the driver. However, abouelenien worked with Mihai burzo, an associate professor of mechanical engineering at the University of Michigan at flint, to try to elevate the concept to several levels. Now, they are developing a system that can automatically detect everyone's "thermal comfort" level, and then constantly adjust the thermal environment. Abouelenien and burzo collected "thermal discomfort" data from 50 subjects and started the experiment. They placed the subjects in a closed environment, roughly simulated the situation of the car compartment, and then recorded various physiological data of the subjects under various temperature conditions. For example, a thermal imaging camera was used to record the facial temperature of the subjects in detail, and three other sensors were used to collect information about respiratory rate, skin temperature and more than 50 other physiological characteristics. At the same time, the subjects will also describe their thermal comfort under various conditions. This is important for abouelenien and burzo to associate such subjective experiences with sensor information and have a concept of "cold" or "hot" from a data perspective. Subsequently, the team created a computer model based on machine learning to create a "decision boundary", that is, to determine the "habitable zone" of a specific individual. On one side of the boundary, the model indicates that the physiological sensor indicates that the person may be feeling cold; on the other side, the model indicates that the person feels too hot. Then, abouelenien and burzo tested the model, and carried out a second experiment, put the subjects under various conditions again, and asked them about their thermal comfort level. As before, the original system will also record the physiological data of the subjects. This time, however, the model was able to use data to predict the response of the subjects, and the results were very shocking. In some cases, especially in cold weather, the model makes a correct prediction of the feelings of the subjects in more than 90% cases. In other words, the model understands how the subjects feel. It's clear that this technology will be applied to car cockpit first, and their algorithm can guide the automatic adjustment of HVAC system, said abouelenien. Another reason that can be used in a car is that the driver or the passenger will keep a relatively constant position, so that the sensor can sense a fixed target. In addition, in addition to comfort, the algorithm guided temperature regulator is more efficient than the driver controlled temperature regulator, and the resulting energy savings may ultimately help improve the range of electric vehicles. As for whether the technology can be applied to homes or offices, abouelenien said it has not yet been implemented. The main challenge is that people often move in such environments, so the system needs a larger sensor network. However, abouelenien also said that it is not difficult to achieve. The researchers found that the price of the thermal imaging camera, the device that best indicates thermal comfort, has fallen a lot. The next goal is to find a way to collect physiological data using non-contact sensors, that is, data can be collected without touching humans. (all pictures are from University of Michigan Dearborn)