the-midnight-paper

Cyclops: Wearable and Single-Piece Full-Body Gesture Input Devices

Authors: Liwei Chan, Chi-Hao Hsieh, Yi-Ling Chen, Shuo Yang, Da-Yuan Huang, Rong-Hao Liang, Bing-Yu Chen

Conference: CHI 2015, Crossings, Seoul, Korea

Full-body gesture input; posture recognition; single-point wearable devices; ego-centric view

Strength

Cyclop’s different placement options allow users to use the device according to the task at hand. This gives users the flexibility of using it to target certain gestures without compromising with the posture. For example, placing it on the chest favor head interactions and placing it around belly enable foot interactions.

Weakness

Excluding female participants to avoid unfavourable factors in evaluation is a major drawback. If some users felt uncomfortable with putting the device on the chest, authors could have suggested placing it on belly and modify the task in pilot study to get more representative results. The claim to support different placement options is inconsistent with their discussions in different sections.

Possible Extension

I would extend Cyclops as an intelligent personal assistant that can not only see and classify user gestures, but also suggest actions based on users’ own blind spots or things what human eyes cannot identify but infrared imaging can(for example: icy slippery surfaces).


Depending on purpose and application, low-level sensing techniques be implemented on certain parts of the body. However, full-body motion input requires a body sensor network of sensors that are distributed on body parts, which may be inconvenient for users to put on. Cyclops overcome this by proposing a single-piece wearable device with wider “field of view”. Cyclops recognizes static and moving bodily gestures based on motion history images (MHI) and a random decision forest (RDF).

Hardware specifications:

Image pre-processing pipeline: Basic idea is to extract foreground images from where the limbs enter the ego-centric view at the edge of the fisheye image. This strategy avoids dealing with non-limb foregrounds in the central part of the image

Relevant work: