Accelerometer Dataset


If you use this dataset, please cite the following paper.  

L. Bao and S. S. Intille, "Activity recognition from user-annotated acceleration data," in Proceedings of PERVASIVE 2004, vol. LNCS 3001, A. Ferscha and F. Mattern, Eds. Berlin Heidelberg: Springer-Verlag, 2004, pp. 1-17.
[Abstract] [PDF]

Here is the link to the data

You can use Winrar to decode it. Its final size is 2.1GB

This is the explanation of the contents:

This CD contains accelerometer data collected using 10G accelerometer hoarder boards. The actlist.mat file contains a cell structure list containing the 20 activities studied.  For example, actlist{1} is “walking”, actlist{2} is “walking while carrying items”, etc.  Actlist{21} is “other”.

Data for each subject is stored in files with naming convention:
agg_s<subject no.>_<activity/obstacle>_<date>.mat,
where <subject no.> is an anonymous identifier for the subject from 1 to 20.  The <activity/obstacle> field specifies whether the data in the file corresponds to laboratory (activity) or semi-naturalistic (obstacle) data. The <date> field corresponds to the date on which the data was collected in MMDDYY format.  For example,
contains semi-naturalistic data for subject number 4 on May 30, 2003.

Each of these data file contains the following MATLAB variables:
actdata{i,j} contains acceleration samples for the activity label actlist{i} from the jth accelerometer source. The indexing order of the sources is as follows: j = 1 corresponds to data from the right hip, 2 is for data from the dominant wrist, 3 is for data from the non-dominant arm, 4 is for data from the dominant ankle, 5 is for data from the non-dominant thigh.
acttime{i,j} contains hoarder timestamps for actdata{i,j}.  There is one timestamp for every 100 samples of accelerometer data.
celldata{j} contains raw acceleration data from accelerometer source j.
celltime{j} contains timestamps for celldata{j}.
acts(i) is the ith activity the subject performed in chronological order during the study.
times(i) specifies the index into celldata{j} of acts(i).  For example, accelerometer data from source j for acts(1) corresponds to data from celldata{j}(times(1):times(2)-1).
freqs(j) contains the average sampling frequency of hoarder source j in Hz.
starttime contains the start time for the study.  The vector specifies the year, month, date, hour, minute, and second in that order the subject began the study.

To illustrate typical use of these variables, here’s two examples.  To draw a spectrogram of x-axis acceleration for all walking data in agg_s4_obstacle_053003.mat, load the file into MATLAB and enter the following at the MATLAB prompt: “figure, spectrogram(actdata{1,1}(1,:))”.  Note actlist{1} is “walking”.  To plot the raw data from the hip accelerometer for the same subject, enter "plot(celldata{1}(1:2,:)')".



Collection of these datasets was made possible by the National Science Foundation and the MIT House_n Consortium. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).