Skip directly to content

listen

SMART WATCHES WILL SOON BE ABLE TO LEARN MUCH MORE ABOUT WHAT YOU ARE DOING

  • johnkelin's picture
    Started by johnkelin
    (21 May '19)

    Researchers from Carnegie Mellon (Carnegie Mellon University) with the help of LG smart watches were able to determine what exactly their owners are doing: using a touch phone, typing or washing their hands.

    Smart watches rafiqsonsonline.com/product-category/smart-watches/ are mainly acquired because they work as fitness trackers, tracking physical activity for the day. Most smart watches are equipped with accelerometers, which measure how movement changes over time, that is, track, walk, run, ride a bike or sleep.

    Recently, two researchers from Carnegie Mellon University using the

    0
    News
on Tue, 2019-05-21 09:50
Forum category: 

Researchers from Carnegie Mellon (Carnegie Mellon University) with the help of LG smart watches were able to determine what exactly their owners are doing: using a touch phone, typing or washing their hands.

Smart watches rafiqsonsonline.com/product-category/smart-watches/ are mainly acquired because they work as fitness trackers, tracking physical activity for the day. Most smart watches are equipped with accelerometers, which measure how movement changes over time, that is, track, walk, run, ride a bike or sleep.

Recently, two researchers from Carnegie Mellon University using the accelerometer in LG smart watches were able to identify the movement of the owners of the watch owners. They made a presentation on this topic at the annual ACM conference on man-machine interaction (Association for Computing Machinery - Association of Computing Machinery).

Thus, electronics can collect user data using a sensor that is already embedded in most modern wearable devices and smartphones.

Scientists tracked the movements of the hands of 50 participants, who noted that they do their hands at regular intervals for almost 1000 hours. This information was used to create a database of the most common hand movements.
The researchers then developed an algorithm that recognizes with an accuracy of 95.2% the subtle differences between the 25 common hand movements, including hand washing, washing dishes, scrolling the screen of a smartphone, using the remote, and typing text.
To capture such subtle differences between movements, the researchers moved the accelerometers to high-speed mode.

This allowed them to obtain more detailed information, including the position of the hand, the patterns of movement, and even bioacoustic data, which are microvibrations that spread up the hand of the owner of the watch. Chris Harrison, head of the Future Interfaces Group at Carnegie Mellon University and co-author of the report, says it’s almost the same as putting a stethoscope on your arm. The convolutional neural network, a type of machine learning algorithm, was able to establish patterns using all the collected information and associate them with certain hand movements.

“This opens up completely new possibilities for recognizing movements with high precision, which was previously impossible,” Harrison wrote in an e-mail to Fast Company magazine. “We can, for example, track how you type in order to recommend (or provide) breaks.”

Harrison points out other similar context-sensitive applications . For example, a watch can track when and for how long you eat, and then pass this information to a calorie counting application. In the same way, a watch can remind you to drink more water if it is recorded that you drank an insufficient amount on that day.

Although these applications open up new possibilities for use, they can also cause privacy issues. For example, this study shows that if Apple implemented a similar system, the company would eventually receive detailed information about what the watch owners are doing most of the time, because the movement of the arms is inextricably linked to our business.

Therefore, the co-author of the report and graduate student Carnegie Mellon Jirad Lapet believes that the whole process of machine learning should occur on the device without connecting to the cloud. This is a key condition for the preservation of personal information of users.

Research shows that it is possible to obtain even more information about how people spend their time using such simple and affordable sensors as an accelerometer. And although these technologies are still in scientific circles, there is a high probability that they will soon appear in your smartwatch, considering how much IT companies are striving to collect data