Walking-Dataset

This dataset is made of 20 IMUs recordings of people walking 5 meters. The IMUs are positioned on both legs in the foot-shank-thigh configuration. The subjects are 20 elderly people and 20 young people.

Dataset Organization

The dataset is set in this way:

• There are 20 recordings of eldery people in the folder \Dataset\Elderly and other 20 recordings of young people in the folder \Dataset\Young, both the groups were asked to walk for 5 meters straight.

• There is \Dataset\Additional folder to which belong:

1. a subfolder Marzia with recordings of me walking 5 meters straight, in a rectagle of 5x3 meters and in a circle of 3.6 meters diameter.
2. a subfolder Long Distance with recordings of me walking randomly in a corridor.
3. a subfolder Disability with 2 recordings of a subject affected by legs muscolar atrophy, walking with sticks and leg braces.

In each folder of \Dataset you can find the recordings in .xlsx format There are 7 sensors and each one corresponds to a part of the body: _1 right foot _2 right shank _3 right thigh _4 left thigh _5 left shank _6 left foot (_7 back, but it is not syncronized with the other sensors and in some recordings is not available)

The useful information are:

• Acc_read_x linear acceleration along x

• Acc_read_y linear acceleration along y

• Acc_read_z linear acceleration along z

• Gyro_read_x angular velocity x

• Gyro_read_y angular velocity y

• Gyro_read_z angular velocity z

• Ext1 pressure toe

• Ext2 pressure heel

Data Manipulation

In order to use the data you have to:

SCALE acc=(acc/10000)*gravity gyro=(gyro/100)*pi/180

FILTER Taking into account that the sampling frequency is 100 apply a low pass filter to remove the noise with the cut off frequency around 3 Hz

ROTATE RIGHT LEG=[0 0 1; 0 1 0; -1 0 0]; LEFT LEG=[0 0 -1; 0 -1 0; -1 0 0];

Additional Data

There is an additional folder \EstimatedData, with each subfolder corresponding to the dataset. Inside each subfolder there are 3 files .mat and 7 files .csv and additional files .fig that are matlab plots

• angularRate: scaled, filtered and rotated;
• specificForce: scaled, filtered and rotated; (it is the pure accelerometer output, gravity included)
• orientation (roll,pitch and yaw): estimated;
• position (x,y,z): estimated;
• linear velocity magnitude: estimated;
• zupt(zero velocity update): it is 1 when the foot is supposed to have zero velocity and 0 when the foot is moving;
• T=gyro(k)'*gyro(k)/sigmaG^2+acc(k)'*acc(k)/sigmaA^2. Check if the test statistics T are below the detector threshold. If so, chose the hypothesis that the system has zero velocity.