The MemGlove detects hand poses and recognizes objects
Hand movements have long been used as a computer interface method, but as reported here, the MemGlove from a team of MIT CSAIL researchers takes things several steps further. This augmented glove can sense hand poses and how it’s applying pressure to an object.
The wearable uses a novel arrangement of 16 electrodes to detect hand position based on resistance, and six fluid filled tubes that transmit pressure depending on how an item is gripped.
An Arduino Due is used to sense these interactions, which pass information on to a computer for processing. Pose verification is accomplished with a Leap Motion sensor. By training neural networks with TensorFlow, the glove is able to identify various hand poses, as well as distinguish between 30 different household things that are grasped.
More details on the MemGlove can be found in the researchers’ paper here.