How can i describe start and end point of a movement data?

Hello,

I have data of 2 people recorded with inertial sensors, which includes phone conversation activity. Data contains data obtained from joints. These data are velocity, acceleration, joint angle.

What I want to do is to detect the start and end points of the movement and record the data in this range.

I need to automate this as I will be doing it for a large number of patients. Each patient started to move at different moments and I need to detect this. For example, the first patient starts the movement immediately, while the second begins approximately at the 105th frame. While the first patient finished the movement around the 50th frame, the second patient finished the movement around the 195th frame.

I need your help as I have limited coding experience. I would be grateful for your any help.

[Mathieu NOE]

hello

you have quite a bunch of data !! have you already make a decision about which ones are the best to pick for evaluating activity vs rest ?

I had a look at both head angular velocity and (translationnal) velocities

while we could more or less see the angular velocity (computed here as the rms of the x,y,z components) change according to your description, the (translationnal) velocities do not match this activity description at least for patient (1) were its velocity does not drop at the end of the record

so it's more a question about how to find the good data first rather than programming thresholds or statements

patient 1

patient 2

clc;clearvars
close all

load('phonecalling.mat')
ci = 2;
time = patient(ci).time;
frames = 1:numel(time);
x = patient(ci).angularVelocity.C1Head_x;
y = patient(ci).angularVelocity.C1Head_y;
z = patient(ci).angularVelocity.C1Head_z;

Head_angularVelocity_rms = sqrt(x.^2+y.^2+z.^2);
Head_angularVelocity_rms2 = smoothdata(Head_angularVelocity_rms,'movmedian',10);

x = patient(ci).velocity.C1Head_x;
y = patient(ci).velocity.C1Head_y;
z = patient(ci).velocity.C1Head_z;

Head_velocity_rms = sqrt(x.^2+y.^2+z.^2);
Head_velocity_rms2 = smoothdata(Head_velocity_rms,'movmedian',10);

subplot(211),
plot(frames,Head_angularVelocity_rms,frames,Head_angularVelocity_rms2);
title('Head angular Velocity');
legend('rms','rms smoothed');
subplot(212),
plot(frames,Head_velocity_rms,frames,Head_velocity_rms2);
title('Head Velocity');
legend('rms','rms smoothed');

[Mert Dogan]

Thank you for your suggestion.

If i didnt understand wrong, I should get a smooth data before analysis and I should select a data or joint for important for task.

I think best joint for the analysis RWrist, and best axis r x or z.

I have dataset which collected from nearly 40 individuals in total on 16 different conditions.

Therefore, it will be difficult to process them. Sorry for my question

"Head_angularVelocity_rms = sqrt(x.^2+y.^2+z.^2);

Head_angularVelocity_rms2 = smoothdata(Head_angularVelocity_rms,'movmedian',10);"

what exactly did we do here?

How can I detect the start and end points?

[Mathieu NOE]

hello

sqrt(x.^2+y.^2+z.^2) is a rms value of the 3 individual coordinates (whatever the data is)

I have followed your suggestion but the jointAngle data for RWrist (and I also computed it's derivative) did not really help me for the job...

clc;clearvars
close all

load('phonecalling.mat')
ci = 2;
time = patient(ci).time;
frames = 1:numel(time);
x = patient(ci).angularVelocity.C1Head_x;
y = patient(ci).angularVelocity.C1Head_y;
z = patient(ci).angularVelocity.C1Head_z;

Head_angularVelocity_rms = sqrt(x.^2+y.^2+z.^2);
Head_angularVelocity_rms2 = smoothdata(Head_angularVelocity_rms,'movmedian',10);

% x = patient(ci).velocity.C1Head_x;
% y = patient(ci).velocity.C1Head_y;
% z = patient(ci).velocity.C1Head_z;
% 
% Head_velocity_rms = sqrt(x.^2+y.^2+z.^2);
% Head_velocity_rms2 = smoothdata(Head_velocity_rms,'movmedian',10);


x = patient(ci).jointAngle.RWrist_x;
y = patient(ci).jointAngle.RWrist_y;
z = patient(ci).jointAngle.RWrist_z;
xdot = gradient(x);
ydot = gradient(y);
zdot = gradient(z);

jointAngle_RWrist_rms = sqrt(x.^2+y.^2+z.^2);
jointAngle_RWrist_rms2 = smoothdata(jointAngle_RWrist_rms,'movmedian',10);

jointAngle_RWrist_dot_rms =  sqrt(xdot.^2+ydot.^2+zdot.^2);
jointAngle_RWrist_dot_rms2 = smoothdata(jointAngle_RWrist_dot_rms,'movmedian',10);

subplot(311),
plot(frames,Head_angularVelocity_rms,frames,Head_angularVelocity_rms2);
title('Head angular Velocity');
legend('rms','rms smoothed');
subplot(312),
plot(frames,jointAngle_RWrist_rms,frames,jointAngle_RWrist_rms2);
title('jointAngle RWrist');
legend('rms','rms smoothed');
subplot(313),
plot(frames,jointAngle_RWrist_dot_rms,frames,jointAngle_RWrist_dot_rms2);
title('jointAngle RWrist (derivative)');
legend('rms','rms smoothed');

[Mert Dogan]

Thanks a lot 🙏🏻 You helped me a lot. So , I am gonna select begin and finish movement manually from motion capture animation.

[Mathieu NOE]

my pleasure and good luck for the future !