@InProceedings{Berlin2012,
  author    = {Berlin, E. and Van Laerhoven, K.},
  title     = {Detecting Leisure Activities with Dense Motif Discovery},
  booktitle = {Proceedings of the ACM Conference on Ubiquitous Computing},
  year      = {2012},
  pages     = {250--259},
  abstract  = {This paper proposes an activity inference system that has
been designed for deployment in mood disorder research,
which aims at accurately and efficiently recognizing selected
leisure activities in week-long continuous data. The approach
to achieve this relies on an unobtrusive and wrist-worn data
logger, in combination with a custom data mining tool that
performs early data abstraction and dense motif discovery to
collect evidence for activities. After presenting the system
design, a feasibility study on weeks of continuous inertial
data from 6 participants investigates both accuracy and execution speed of each of the abstraction and detection steps.
Results show that our method is able to detect target activities in a large data set with a comparable precision and recall
to more conventional approaches, in approximately the time
it takes to download and visualize the logs from the sensor.},
  acmid     = {2370257},
  doi       = {10.1145/2370216.2370257},
  file      = {Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:Research\\2013-09 SegLitReview\\2013-09\\1 - InJabRef\\Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:PDF},
  groups    = {Lit Review 2013-09, IROS2014},
  isbn      = {978-1-4503-1224-0},
  keywords  = {activity detection, motif discovery, psychiatric monitoring},
  location  = {Pittsburgh, Pennsylvania},
  numpages  = {10},
  review    = {Want to monitor psychiatric patients by examining their movements and mood and using their diaries and activity labels.Actions are recognitized at motifs, and placed into a suffic tree. Motifs are discovred at training time by extracting features found in raw accel peaks and placing them into a suffix tree. Classification done with bag-of-words.

The key factors to consider are: supervised learning (entries from diaries), lon term monitoring (week long, 24/7), small activity set. Want to be able to upload this data once a week so psyciatrists can examine the movement profile of the patient, and can be used to suppliment existing psych scales and questionaires.

They build a watch that logs accel, light and temperature at 100 Hz. Linear segments are created from observation data (piecewise linear approximation - Keogh2001), using online alg that minimizes residual error between linear segment and obs data.They look at the slope and length of each linear segment and assign a symbol to it. The slope of the linear segment was converted to an angle, and binned. Low motion (slope close to 0) were assumed to be stationary data and not considered. The labels were assigned by taking these bins and assign labels based on each two subsequent motions and their angle bin.This creates a long sequence of symbols, which can be searched for a motiff by suffix tree. Bag of words used to find occurances of events once the suffix are labeled.

Berlin and Van Laerhoven \cite{Berlin2012} monitor psychiatric patients using accelerometers on a wrist watch, and applied Keogh's performed piecewise linear approximation \cite{Keogh2004}. The slope of the linear segments are converted to angles and binned, thus discretizing the data. Symbols are assigned to sequential pairs of bins (see Figure \ref{fig:Berlin2012_binAngles}). Motifs are located from these symbols by a suffix tree, and labelled using a bag-of-words classifier. 6 subjects were assessed, over the span of a day. In general, the proposed approach outperforms SVM trained with mean, variance and FFT features.},
  timestamp = {2013.09.29},
}

{"_id":"ZwZH2HZxhTgYWAveo","bibbaseid":"berlin-vanlaerhoven-detectingleisureactivitieswithdensemotifdiscovery-2012","downloads":0,"creationDate":"2017-09-14T16:34:36.214Z","title":"Detecting Leisure Activities with Dense Motif Discovery","author_short":["Berlin, E.","Van Laerhoven, K."],"year":2012,"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/jfslin/jfslin.github.io/master/jf2lin.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["Berlin"],"firstnames":["E."],"suffixes":[]},{"propositions":[],"lastnames":["Van","Laerhoven"],"firstnames":["K."],"suffixes":[]}],"title":"Detecting Leisure Activities with Dense Motif Discovery","booktitle":"Proceedings of the ACM Conference on Ubiquitous Computing","year":"2012","pages":"250--259","abstract":"This paper proposes an activity inference system that has been designed for deployment in mood disorder research, which aims at accurately and efficiently recognizing selected leisure activities in week-long continuous data. The approach to achieve this relies on an unobtrusive and wrist-worn data logger, in combination with a custom data mining tool that performs early data abstraction and dense motif discovery to collect evidence for activities. After presenting the system design, a feasibility study on weeks of continuous inertial data from 6 participants investigates both accuracy and execution speed of each of the abstraction and detection steps. Results show that our method is able to detect target activities in a large data set with a comparable precision and recall to more conventional approaches, in approximately the time it takes to download and visualize the logs from the sensor.","acmid":"2370257","doi":"10.1145/2370216.2370257","file":"Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:Research\\\\2013-09 SegLitReview\\\\2013-09\\\\1 - InJabRef\\\\Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:PDF","groups":"Lit Review 2013-09, IROS2014","isbn":"978-1-4503-1224-0","keywords":"activity detection, motif discovery, psychiatric monitoring","location":"Pittsburgh, Pennsylvania","numpages":"10","review":"Want to monitor psychiatric patients by examining their movements and mood and using their diaries and activity labels.Actions are recognitized at motifs, and placed into a suffic tree. Motifs are discovred at training time by extracting features found in raw accel peaks and placing them into a suffix tree. Classification done with bag-of-words. The key factors to consider are: supervised learning (entries from diaries), lon term monitoring (week long, 24/7), small activity set. Want to be able to upload this data once a week so psyciatrists can examine the movement profile of the patient, and can be used to suppliment existing psych scales and questionaires. They build a watch that logs accel, light and temperature at 100 Hz. Linear segments are created from observation data (piecewise linear approximation - Keogh2001), using online alg that minimizes residual error between linear segment and obs data.They look at the slope and length of each linear segment and assign a symbol to it. The slope of the linear segment was converted to an angle, and binned. Low motion (slope close to 0) were assumed to be stationary data and not considered. The labels were assigned by taking these bins and assign labels based on each two subsequent motions and their angle bin.This creates a long sequence of symbols, which can be searched for a motiff by suffix tree. Bag of words used to find occurances of events once the suffix are labeled. Berlin and Van Laerhoven i̧teBerlin2012 monitor psychiatric patients using accelerometers on a wrist watch, and applied Keogh's performed piecewise linear approximation i̧teKeogh2004. The slope of the linear segments are converted to angles and binned, thus discretizing the data. Symbols are assigned to sequential pairs of bins (see Figure e̊ffig:Berlin2012_binAngles). Motifs are located from these symbols by a suffix tree, and labelled using a bag-of-words classifier. 6 subjects were assessed, over the span of a day. In general, the proposed approach outperforms SVM trained with mean, variance and FFT features.","timestamp":"2013.09.29","bibtex":"@InProceedings{Berlin2012,\n author = {Berlin, E. and Van Laerhoven, K.},\n title = {Detecting Leisure Activities with Dense Motif Discovery},\n booktitle = {Proceedings of the ACM Conference on Ubiquitous Computing},\n year = {2012},\n pages = {250--259},\n abstract = {This paper proposes an activity inference system that has\nbeen designed for deployment in mood disorder research,\nwhich aims at accurately and efficiently recognizing selected\nleisure activities in week-long continuous data. The approach\nto achieve this relies on an unobtrusive and wrist-worn data\nlogger, in combination with a custom data mining tool that\nperforms early data abstraction and dense motif discovery to\ncollect evidence for activities. After presenting the system\ndesign, a feasibility study on weeks of continuous inertial\ndata from 6 participants investigates both accuracy and execution speed of each of the abstraction and detection steps.\nResults show that our method is able to detect target activities in a large data set with a comparable precision and recall\nto more conventional approaches, in approximately the time\nit takes to download and visualize the logs from the sensor.},\n acmid = {2370257},\n doi = {10.1145/2370216.2370257},\n file = {Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:Research\\\\2013-09 SegLitReview\\\\2013-09\\\\1 - InJabRef\\\\Berlin2012 - Detecting leisure activities with dense motif discovery.pdf:PDF},\n groups = {Lit Review 2013-09, IROS2014},\n isbn = {978-1-4503-1224-0},\n keywords = {activity detection, motif discovery, psychiatric monitoring},\n location = {Pittsburgh, Pennsylvania},\n numpages = {10},\n review = {Want to monitor psychiatric patients by examining their movements and mood and using their diaries and activity labels.Actions are recognitized at motifs, and placed into a suffic tree. Motifs are discovred at training time by extracting features found in raw accel peaks and placing them into a suffix tree. Classification done with bag-of-words.\n\nThe key factors to consider are: supervised learning (entries from diaries), lon term monitoring (week long, 24/7), small activity set. Want to be able to upload this data once a week so psyciatrists can examine the movement profile of the patient, and can be used to suppliment existing psych scales and questionaires.\n\nThey build a watch that logs accel, light and temperature at 100 Hz. Linear segments are created from observation data (piecewise linear approximation - Keogh2001), using online alg that minimizes residual error between linear segment and obs data.They look at the slope and length of each linear segment and assign a symbol to it. The slope of the linear segment was converted to an angle, and binned. Low motion (slope close to 0) were assumed to be stationary data and not considered. The labels were assigned by taking these bins and assign labels based on each two subsequent motions and their angle bin.This creates a long sequence of symbols, which can be searched for a motiff by suffix tree. Bag of words used to find occurances of events once the suffix are labeled.\n\nBerlin and Van Laerhoven \\cite{Berlin2012} monitor psychiatric patients using accelerometers on a wrist watch, and applied Keogh's performed piecewise linear approximation \\cite{Keogh2004}. The slope of the linear segments are converted to angles and binned, thus discretizing the data. Symbols are assigned to sequential pairs of bins (see Figure \\ref{fig:Berlin2012_binAngles}). Motifs are located from these symbols by a suffix tree, and labelled using a bag-of-words classifier. 6 subjects were assessed, over the span of a day. In general, the proposed approach outperforms SVM trained with mean, variance and FFT features.},\n timestamp = {2013.09.29},\n}\n\n","author_short":["Berlin, E.","Van Laerhoven, K."],"key":"Berlin2012","id":"Berlin2012","bibbaseid":"berlin-vanlaerhoven-detectingleisureactivitieswithdensemotifdiscovery-2012","role":"author","urls":{},"keyword":["activity detection","motif discovery","psychiatric monitoring"],"downloads":0},"search_terms":["detecting","leisure","activities","dense","motif","discovery","berlin","van laerhoven"],"keywords":["activity detection","motif discovery","psychiatric monitoring"],"authorIDs":[],"dataSources":["iCsmKnycRmHPxmhBd"]}