10/02/2016
Gesture Recognition Unit (GRU) - Exoglove
From November 2008 to March 2010, RallyPoint developed the Gesture Recognition Unit (GRU) exoglove as a subcontractor for Lockheed Martin's Collaborative Squad Tactical Augmented Reality (COSTAR) wearable soldier system under DARPA’s Urban Leader Tactical Response, Awareness & Visualization (ULTRA-Vis) program. Key advantages of the GRU with the intended application include:
Compatible with gloves and bare hand:
The GRU is based on a rugged "exoglove" platform that can be worn over existing combat-proven protective Nomex gloves or on a bare hand. It has minimal coverage at the fingers and palm areas, resulting in a much better feel than other forms of over- or under-gloves that create excessive bulk. When the Nomex glove wears out, the user can switch to new gloves without having to replace the GRU exoglove. The abrasion- and flame-resistant fabric of the exoglove provides additional protection to the user.
Functional even when heavily soiled or submerged:
A significant portion of input gloves utilizes touch sensors based on capacitance- sensing or electrical contact. These gloves may be considered "water-proof" because they can be washed in water without damaging the components, but they do not function reliably when wet, submerged, or in contact with conductive materials. Furthermore, these sensors rely on clean physical contact, which can be hindered by heavy soiling of the glove. In contrast, the GRU buttons work reliably even when the glove is submerged in water or covered with mud.
Reassuring with tactile feedback:
The GRU buttons provide a satisfying click when pressed. Operators prefer this type of localized tactile feedback over audio and vibration cues, which give imprecise feedback because the audio or vibration sources are typically located a distance away from the finger contact points.
Robust against false-activations:
The GRU buttons are recessed, thus unaffected when bumped against flat, hard surfaces. The buttons are easy to click with the thumb, but do not falsely activate when the user squeezes his fingers together or manipulates various objects.
Fast and Intuitive:
The push-to-gesture button enables the GRU user to have considerable control over the gesture input process. As a result, gestures are recognized rapidly with minimal false-alarms. Furthermore, the GRU enables in-air freehand mouse pointing. This mode of pointing––reminiscent of using a laser pointer or Nintendo Wii controller––is quick and intuitive, and opens doors to additional capabilities such as virtual graffiti and handwriting recognition.
Robust against unintended body motions:
The typical gesture recognition algorithm looks at unprocessed data from individual accelerometers, which are heavily influenced by translational motions of the user's body. Consequently, these algorithms cannot be reliably used to recognize gestures performed when the user is running (especially during rapid accelerations and decelerations) or riding a vehicle on bumpy terrain. The GRU's gesture recognition algorithm is more robust against body motions because it looks primarily at the lateral-motion-independent orientations (Euler angles derived from quaternion representations) of its hand and shoulder modules. These angles are determined by a matrix approach using data from accelerometers and gyroscopes. As a result, we are able to demonstrate reliable recognition of gestures performed when a user is jumping up and down, and making sudden jerky movements in various directions.
