Timothy Campbell and Jonathan Harper and Björn Hartmann and Eric Paulos

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2015-172

July 15, 2015

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-172.pdf

Online communities for sharing instructional content have grown from a renewed interest in DIY culture. However, it is difficult to convey the tacit knowledge implicit in certain skills. We identify the need for Digital Apprenticeship, where workshop activities are sensed and analyzed for both quantitative and qualitative measures. We evaluated this concept with an activity recognition system for carpentry tools. Using a single ring-worn inertial measurement unit (IMU), we collected data from 15 participants using 5 hand and power tools. Our window-based multi-class SVM achieves 82% accuracy with realistic training scenario and outputs user-friendly event activity. We investigate how these results contextualize to applications in digital apprenticeship, namely tutorial authoring, content following and technique feedback.

---

BibTeX citation:

@techreport{Campbell:EECS-2015-172,
    Author= {Campbell, Timothy and Harper, Jonathan and Hartmann, Björn and Paulos, Eric},
    Title= {Towards Digital Apprenticeship: Wearable Activity Recognition in the Workshop Setting},
    Year= {2015},
    Month= {Jul},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-172.html},
    Number= {UCB/EECS-2015-172},
    Abstract= {Online communities for sharing instructional content have grown from a renewed interest in DIY culture. However, it is difficult to convey the tacit knowledge implicit in certain skills. We identify the need for Digital Apprenticeship, where workshop activities are sensed and analyzed for both quantitative and qualitative measures. We evaluated this concept with an activity recognition system for carpentry tools. Using a single ring-worn inertial measurement unit (IMU), we collected data from 15 participants using 5 hand and power tools. Our window-based multi-class SVM achieves 82% accuracy with realistic training scenario and outputs user-friendly event activity. We investigate how these results contextualize to applications in digital apprenticeship, namely tutorial authoring, content following and technique feedback.},
}

EndNote citation:

%0 Report
%A Campbell, Timothy 
%A Harper, Jonathan 
%A Hartmann, Björn 
%A Paulos, Eric 
%T Towards Digital Apprenticeship: Wearable Activity Recognition in the Workshop Setting
%I EECS Department, University of California, Berkeley
%D 2015
%8 July 15
%@ UCB/EECS-2015-172
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-172.html
%F Campbell:EECS-2015-172