Chapter 2 Background
This chapter presents research in related elds to the work conducted for this thesis. Several key concepts from previous academic research e orts as well as commercial products were leveraged to develop the embedded system design approach presented in this thesis. Infor-mation from these areas was also used for the development of the electronic textile prototype platform on which experimentation was performed.
Electronic Textiles and Wearable Computing
Textiles, in general, are networks of threads or yarns that are woven in intersecting patterns to create a exible fabric substrate. Electronic textiles (e-textiles) are simply fabrics with the addition of materials with electrical characteristics. These electrical materials typically include electronics for data processing as well as an assortment of wires or lms used for sensing, communication, power distribution, and actuation. A eld closely related to e-textiles is wearable computing in which computing elements are attached or worn on the body for a variety of purposes such as health monitoring, communication, or even more recently for fashion.
In e-textiles and wearable computing research conducted thus far, particular attention has been focused in the medical eld as well the development of systems for military applications. Projects like Georgia Tech’s wearable motherboard  and the US Army Soldier Systems Center’s smart vest  are excellent examples of the current state of wearable computing. Although medical and military applications have been the major driving force thus far for e-textiles, it is believed entertainment and personal safety will be next elds to push e-textiles and wearable computing into the mainstream. This shift in elds is not because entertainment and personal safety products will feature the most cutting edge sophisticated technology, but rather because they incorporate technologies and purposes geared toward everyday activities. Products like PDD’s Illuminated Cycling Jacket , Swany’s G.Cell hands free cell phone ski and snowboard glove , Raymarine’s LifeTag man-overboard safety communicator , and the abundance of iPod related clothing from companies like Zegna , Burton , and Nike  support this claim.
These entertainment and safety systems will need to follow a di erent set of criteria in order to become widely adopted and to be e ective for use in everyday life. One of a kind, custom, and expensive systems will not su ce. As such, the next generation of e-textiles must feature low power consumption, low material costs, fault tolerance, and must be easily manufactured through some automated process for cost-e ective mass distribution.
Previous Electronic Textile Research at Virginia Tech
The Virginia Tech E-Textiles Laboratory has conducted many research e orts in the eld of electronic textiles in the past several years. Many of these projects have particular focus in wearable computing as well as large-scale sensor networks. The Virginia Tech E-Textiles Laboratory utilizes a computer controlled, automated loom to weave custom smart fabrics in-house using standard textile manufacturing techniques. Although most e-textile research e orts performed at Virginia Tech are mainly proof of concept in nature, particular emphasis is placed on design methodologies for mass-manufacturing and fault-tolerance reliability.
Some examples of e-textiles previously developed at Virginia Tech include an acoustic beam-former , a smart carpet , and an activity recognition pants system . Important design methodologies can be leveraged from each one of these projects. For example, the acoustic beamformer consisted of a 30-foot long computational fabric capable of determining the location and movement of approaching vehicles such as tanks and trucks. Acoustical sens-ing devices as well as computing elements were distributed across the textile substrate and interconnected to form a communications network. A software simulator was also created to aid in the development of the beamformer prototype.
In contrast to the military-oriented application of the beamformer, other projects like the smart carpet serve as excellent platforms for commercial and safety applications. As part of a joint research venture between Virginia Tech and the Intel Corporation, a oor mounted carpet electronic textile was developed with various sensing and actuation technologies . Using an intersecting grid of resistive wires and rows of parallel peizo-electric cables, both a software simulation model and hardware prototype system were constructed to support a footstep tracking application. The smart carpet also featured visual actuation through the use of electro-luminescent (EL) wires. An important design feature of the smart carpet was the use of a publish-subscribe service protocol. Utilizing the concept of services allows developers to work at a higher level of abstraction and helps to facilitate the generation of more complex applications.
The activity recognition pants system  is perhaps the most in uential and relevant e ort from Virginia Tech with regard to the research presented in this thesis. The pants system essentially consisted of a wearable lower body textile substrate with several acceleration modules symmetrically distributed on the hips, knees, and ankles. Piezo-electric lms were also included to capture heel strike data. Raw 2-d motion data was streamed directly o of the pants using a serial Bluetooth device connected to a backend PC. The backend PC executes an activity recognition application which processes the raw motion data. Using Singular Value Decomposition (SVD), the classi cation environment is able to e ectively match the movements of a person wearing the pants to a corresponding set of trained motions in pseudo-realtime.
Sensor Based Activity Recognition
Sensor based activity recognition has been gaining much popularity, especially in the en-tertainment and consumer electronics industries. The Nintendo Wii game console utilizes a hand-held controller device containing a 3-d accelerometer along with a small infrared camera to provide user input to game applications using a pre-determined range of hand gestures and movements . The Sony Playstation 3 (PS3) also features an accelerom-eter based tilt-sensing controller to provide input to the console for motion-based gaming activities . Other more discrete consumer electronics devices like laptop computer hard drives use sensors to detect when the object is free falling with respect to gravity . Many consumer products use sensors to recognize speci c activities, like the Nike iPod shoe inserts which use piezo-electric sensors to monitor running statistics .
1.3 Thesis Organization
2.1 Electronic Textiles and Wearable Computing
2.2 Previous Electronic Textile Research at Virginia Tech
2.3 Sensor Based Activity Recognition
3 System Overview
3.1 Design Hierarchy and Data Flow
3.2 Two-Tier Model
4 Hardware Development
4.1 Tier 1 Acceleration Modules
4.2 Tier 2 Processing Element
4.3 Interconnections and Textile Substrate
5 Software Development
5.1 Tier 1 Firmware
5.2 Tier 2 Software
7.1 Individual Acceleration Module Accuracy
7.2 Activity Recognition Application Performance
7.3 Network Communication Performance
8.2 Future Work
GET THE COMPLETE PROJECT