Working out with free weights is an excellent way to build strength and burn off fat. But keeping track of the sets and reps can be tedious. It’s easy to accidentally skip steps, miscount a sequence, or forget your progress.

Magic Hand
Microsoft Research Cambridge

The new technique uses sensor-embedded gloves and a waistband designed to recognize what type of exercise a person is doing and how many repetitions have been completed. Tied with an online community of exercisers, the system could lead to a new kind of digital personal trainer that gives real-time tips or warnings on form and posture and also connects people with others to help achieve their goals.

"People can use this to share their progress with others. They can share their secret and how they do their exercise," said Keng-hao Chang, a graduate student of computer science at the University of California, Berkeley.

Workout gloves are fixed with wireless sensors called accelerometers, which track motion in three directions: side to side, up and down, and front and back. The data collected is sent via a Bluetooth connection to a computer, where custom software analyzes the information to distinguish a bicep curl from a tricep curl. (Eventually the data could be sent to a mobile phone or PDA.)

Another accelerometer on the belt also track motions in three axes to distinguish whether the exerciser is standing or lying on a bench. That information is helpful because the motion of some exercises — for example a bench press, which works the chest, and the movement of an overhead dumbbell press, which works the shoulders — are quite similar.

In initial tests, the sensors were between 85 percent and 95 percent accurate in recognizing the exercise being performed. And out of 100 repetitions, the system miscounted by less than five.

Chang and his team think that the system could be further enhanced by attaching radio frequency tags to the weights themselves. That information could be picked up by the sensors to determine how much weight is being lifted. In this way, the exerciser need only focus on the workout instead of its progress.

"The application itself is very interesting," said Jamie Ward, a researcher of human activity recognition at Lancaster University in Lancaster, England. "I believe you could make a product out of it."

But before that happens, he said, "There are a number challenges to overcome first."

For example, the sensors need battery power.

"It would be nice if the devices could be powered just from he movements themselves," said Ward. "Wristwatches can do that but they use a very small amount of power, and accelerometers use more power."

There is also the issue of privacy.

"If the environment is smart so that the objects are monitoring you, it obviously raises some privacy issues," said Ward.

It will be important to keep the information localized, he said.

Via: Discovery Channel