BWR-MultiDevice

Dataset Overview:

The BWR-MultiDevice Dataset collects gait templates from 25 subjects acquired in 2 sessions at time distance of about 15 days.

Each session is made up by 6 walking signals registered with 3 different kinds of devices (a OnePlus One with a LIS3DH accelerometer, a Samsung Galaxy S4 Active with a K330 3-axes accelerometer, and a Sony Xperia S with a BMA250 accelerometer), 2 for each device. The acquisition device is positioned in the right or left hip attached to a belt.

After each capture, the device is detached in order to introduce further variations.

Each walking signal is collected along a straight hallway and contains 10 steps for a total of about 1300 samples per walk.

This dataset provides both original data and normalized ones (with the procedure described in detail in the related paper - see below).

Archive Structure:

- BWR-MultiDevice Dataset

|

|___Original Dataset

| |___subj01

| | |___walk01

| | | |___SonyXperiaS_subj01_S1_C1.csv

| | |___ ...

| | |___walk12

| | | |___SamsungS4_subj01_S2_C2.csv

| ..........................

| |___subj25

| |___walk01

| | |___SonyXperiaS_subj01_S1_C1.csv

| |___ ...

| |___walk12

| | |___SamsungS4_subj01_S2_C2.csv

|

|___Normalized Dataset

| |___subj01

| | |___walk01

| | | |___SonyXperiaS_subj01_S1_C1.csv

| | |___ ...

| | |___walk12

| | | |___SamsungS4_subj01_S2_C2.csv

| ..........................

| |___subj25

| |___walk01

| | |___SonyXperiaS_subj01_S1_C1.csv

| |___ ...

| |___walk12

| | |___SamsungS4_subj01_S2_C2.csv

|

|___SubjectData.csv

Data Details

For each subject:

walk01-walk02 folders: Contain data collected with Sony Xperia S in the 1° session
walk03-walk04 folders: Contain data collected with OnePlus One in the 1° session
walk05-walk06 folders: Contain data collected with Samsung S4 Active in the 1° session
walk07-walk08 folders: Contain data collected with Sony Xperia S in the 2° session
walk09-walk10 folders: Contain data collected with OnePlus One in the 2° session
walk11-walk12 folders: Contain data collected with Samsung S4 Active in the 2° session

Each walk signal is stored as a .csv file with 4 columns separated by ";" with labels "time", "x", "y", and "z" , and as many rows as the number of samples.

According to Android capture standard, data is not time interpolated. Samples are recorded on "significant" variations (see Android standard).

The file SubjectData.csv collects the information about gender and age of the 25 subjects.

Further Details:

The table below describes the age/gender distribution.

Usage Request:

The database is protected by password.

Before we can send you the password, we need a declaration from the director of the interested Laboratory (even via e-mail from his/her institutional e-mail address) in which he/she declares that the database will be used only for scientific purposes and that the source paper will be cited in any publication in which these data will be used.

A template can be found here.

Maria De Marsico, Daniele De Pasquale, and Alessio Mecca. Embedded accelerometer signal normalization for cross-device gait recognition. A. Bromme, C. Busch, C. Rathgeb and A. Uhl (Eds.): BIOSIG 2016, Lecture Notes in Informatics (LNI), Gesellschaft fur Informatik, Bonn 2016

Requests must be addressed to either Alessio Mecca (mecca@di.uniroma1.it) or Maria De Marsico (demarsico@di.uniroma1.it)

Download Link:

BWR-MultiDevice Dataset