Data Set

Neura Sparse Dataset

Setup

The experiment had nine healthy subjects (7 men and 2 women, weight \(63.0 \pm 6.8\) kg, height \(1.70 \pm 0.06\) m, age \(24.6 \pm 3.9\) years old), with no known gait or lower body biomechanical abnormalities. Each subject was compared to two benchmark systems, namely the Vicon and Xsens systems.

  • The Vicon Vantage system consisted of eight cameras covering approximately \(4 \times 4 m^2\) capture area with millimetre accuracy. Vicon data were captured at 100 Hz and processed using Nexus 2.7 software.

  • The Xsens Awinda system consisted of seven MTx units (IMUs). Xsens data were captured at 100 Hz using MT Manager 4.8 and processed using MVN Studio 4.4 software.

The Vicon and Xsens recordings were synchronized by having the Xsens Awinda station send a trigger pulse to the Vicon system at the start and stop event of each recording. Each subject had reflective Vicon markers placed according to the Helen-Hayes 16 marker set, seven MTx units attached to the pelvis, thighs, shanks, and feet according to standard Xsens sensor placement, and two MTx units attached near the ankles. The MTx units were all factory calibrated.

Refer to the following papers for examples on how the dataset can be used: [1] [2].

Movements

Each subject performed the movements listed in the table below twice (i.e., two trials). The subjects stood still before and after each trial for ten seconds. The experiment was approved by the Human Research Ethics Board of the University of New South Wales (UNSW) with approval number HC180413.

Movement

Description

Duration (s)

Static

Stand still

~10

Walk

Walk straight and back

~30

Figure of eight

Walk in figures of eight

~60

Zig-zag

Walk zigzag

~60

5-minute walk

Undirected walk, side step, and stand

~300

Speedskater

Speedskater on the spot

~30

Jog

Jog straight and return

~30

Jumping jacks

Jumping jacks on the spot

~30

High knee

High knee jog straight and return

~30

Installation

  1. Download neura-sparse01.zip from Harvard Dataverse. Refer to Description (base) for details.

  2. Extract to data/neura-sparse01

  3. If you want the raw recordings (not time aligned, and no calibration done), download neura-sparse01-raw.zip from the link above. Refer to Description (raw) for details.

Description (base)

<Subject ID>-Trial-<Movement Type>.<Extension>

Note that imu, vicon, and step-detect are all aligned (time-wise), and are with respect the world frame.

imu

File name is <Subject ID>-Trial-<Movement Type>-<BodyPart>.csv.

Body Segment

Sensor Location

Pelvis

Pelvis backside

L_UpLeg

Left thigh

R_UpLeg

Right thigh

L_LowLeg

Left ankle (slightly above)

R_LowLeg

Right ankle (slightly above)

L_LowLeg2

Left shanks

R_LowLeg2

Right shanks

L_Foot

Left foot (w/ shoes)

R_Foot

Right foot (w/ shoes)

Column Description:

Column

Type

Description

Acc_X/Y/Z

Float

Measured acceleration in sensor frame

Gyr_X/Y/Z

Float

Measured angular velocity in sensor frame

Mag_X/Y/Z

Float

Measured magnetic field in sensor frame

ori_q0-4

Float

Orientation of sensor in world frame (w,x,y,z)

Example:

PacketCounter,SampleTimeFine,Acc_X,Acc_Y,Acc_Z,Gyr_X,Gyr_Y,Gyr_Z,Mag_X,Mag_Y,Mag_Z,Quat_q0,Quat_q1,Quat_q2,Quat_q3
27626,,8.696773,0.341363,4.715198,0.015417,0.009366,-0.011596,0.886719,0.420898,0.468506,0.651925,-0.318177,-0.399341,-0.560611
...

vicon

CSV export of ViconBody. Contains marker and joint centre positions of the movement trial.

xsens

Pose reconstruction by the Xsens system. Specifically, bvh files of the movement trial generated by the MVN Studio 4.4 software. Note that its index are not aligned with vicon and imu (haven’t edited them yet). Data loader code for imu should return the corresponding index that must be used. This data is for comparing with the xsens MVN black box output.

step-detect

Indicate if step is detected. Reviewed manually.

Column

Type

Description

stepL

Boolean

1 if left foot step is detected, 0 if not

stepR

Boolean

1 if left foot step is detected, 0 if not

Example:

stepL,stepR
1,1
...

calib

  • <Subject ID>-Trial-<Movement>-Calib-SensorYawFixWorldFrame.txt: Contains yaw offset calibration for pelvis, ankle, and foot IMUs. Only important file in the folder.

  • <Subject ID>-Calib-V2W-Compass.mat and <Subject ID>-Calib-V2W-Pendumum.mat: used to calculate Vicon to World rotation matrix. Only used in raw processing.

  • <Subject ID>-Calib-W2V.txt: Description to follow. Did not use in the dataset.

Description (raw)

rawvicon

CSV export from vicon but converted to .mat file. Contains marker and joint centre positions of the movement trial.

rawimu

File name is <Subject ID>-Trial-<Movement Type>-<Sensor ID>.txt.

Sensor ID to body segment table:

Body Segment

Sensor ID

Pelvis

00B40B91

L_UpLeg

00B40C45

R_UpLeg

00B40C3C

L_LowLeg

00B40C44

R_LowLeg

00B40C47

L_LowLeg2

00B40BA5

R_LowLeg2

00B40C35

L_Foot

00B40C55

R_Foot

00B40C48

Column Description:

Column

Type

Description

PacketCounter

Int

Packet number

SampleTimeFine

Float

Time of recording

Acc_X/Y/Z

Float

Measured acceleration in sensor frame

Gyr_X/Y/Z

Float

Measured angular velocity in sensor frame

Mag_X/Y/Z

Float

Measured magnetic field in sensor frame

Quat_q0-4

Float

Orientation of sensor in world frame

Example:

PacketCounter,SampleTimeFine,Acc_X,Acc_Y,Acc_Z,Gyr_X,Gyr_Y,Gyr_Z,Mag_X,Mag_Y,Mag_Z,Quat_q0,Quat_q1,Quat_q2,Quat_q3
27626,,8.696773,0.341363,4.715198,0.015417,0.009366,-0.011596,0.886719,0.420898,0.468506,0.651925,-0.318177,-0.399341,-0.560611
...

rawstep-detect

Indicate if step is detected for the whole raw trial. Reviewed manually.