Acceleration Sensor Gait Dataset
----------------------------
(c) 2012, Jaakko Suutala

Please cite [1] when using the dataset.

Datasets
---------
Each dataset file contains feature vectors as rows and each column gives the 
values of that particular feature, first column being order number of sequence 
(NOT USED AS FEATURE!) and last column being a true class label (i.e., person id). 
If you want to classify single feature vectors forget the first column. You can 
combine these to floor measurements in another datasets, which were collected simultaneously.
100Hz sampling rate was used and the 3D accelerometer was attach to person's necklace
(cf. [1] for details).

Induvidual datasets are:

9dataset_acc.txt: 25 bin FFT amplitude spectrum features from walking sequence using 50 sample sliding 
Hanning window with 50% overlap, 9 subjects and 20 walking seq. / subject

4dataset_acc.txt: 25 bin FFT amplitude spectrum features from walking sequence using 50 sample sliding 
Hanning window with 50% overlap, 4 subjects and ca. 80 walking seq. / subject. 3 different footwear and 
walking speed collected from each subject

9dataset_acc_all.txt: combined dataset of two presented above. Totally 418 walking sequences 
from 9 subjects  


References
----------
[1] Suutala J. (2012) Learning Discriminative Models from Structured 
Multi-sensor Data for Human Context Recognition, Doctoral Thesis, University of Oulu.