There is an error in the part that searches for the player and the bodypart int the .tbm file. It's fixed in this one.
Last edited by psycore; Dec 8, 2011 at 03:36 AM.
Reason: New error. Must have been drunk or something oO
Signature temporarily out of order.
Getting joint rotation
EDIT: Sry, i'm an idiot. This message should be in this thread.
Is there any way to get joint rotation among each other better than some magic manipulations with body_info.rot or counting them from joint world coods?
I'm just trying to do .rpl - .bvh converter, and made basic script, that saves replays as array of POS data of every frame. But .bvh uses only rotation of joints, so this script is completely unuseful.
I see only two ways how to get joint rotation:
1. Just subtract world angles of joint, for example, rotation of elbow among shoulder would be something like rot.elbow=elbow.y - shoulder.y, it cannot be rotated any other angle, so it should be simple.
The problem with it is that i dont understand, how body_info.rot works. Only one match in google was this thread with some strange code with many glitches.
2. Calculate rotation thru arcsin. For example, if x,y,z=elbow[x,y,z]-shoulder[x,y,z], then rot.z=arcsin(y/((x^2+y^2)^.5)). The only problem is too much operations, and, as a consequence, low performance.
Any ideas how to do it?
Is there any way to get joint rotation among each other better than some magic manipulations with body_info.rot or counting them from joint world coods?
I'm just trying to do .rpl - .bvh converter, and made basic script, that saves replays as array of POS data of every frame. But .bvh uses only rotation of joints, so this script is completely unuseful.
I see only two ways how to get joint rotation:
1. Just subtract world angles of joint, for example, rotation of elbow among shoulder would be something like rot.elbow=elbow.y - shoulder.y, it cannot be rotated any other angle, so it should be simple.
The problem with it is that i dont understand, how body_info.rot works. Only one match in google was this thread with some strange code with many glitches.
2. Calculate rotation thru arcsin. For example, if x,y,z=elbow[x,y,z]-shoulder[x,y,z], then rot.z=arcsin(y/((x^2+y^2)^.5)). The only problem is too much operations, and, as a consequence, low performance.
Any ideas how to do it?
Last edited by rfxDarth; Jan 4, 2012 at 07:47 AM.
HAMPA SUX!
URL removed ~SF
URL removed ~SF
The body rotation just tells you how the bodypart is rotated around its center, not around its joint.
If what you wrote in 2. already has to much operations I don't see much you can do. I'm not sure how the "^" operator works but it may be faster if you just write x*x instead of x^2.
I don't quite get which angle you're trying to get. I made a little sketch and it doesn't seem to make any sense:
I assume that you want the angle between biceps and triceps to measure how far the elbow is rotated. To get that you'll need far more operations than those you wrote.
Are you calculating the coordinates while playing the replay? If yes then I'd reccomend calculating them first, saving them into an array and the use the saved values.
If what you wrote in 2. already has to much operations I don't see much you can do. I'm not sure how the "^" operator works but it may be faster if you just write x*x instead of x^2.
I don't quite get which angle you're trying to get. I made a little sketch and it doesn't seem to make any sense:
I assume that you want the angle between biceps and triceps to measure how far the elbow is rotated. To get that you'll need far more operations than those you wrote.
Are you calculating the coordinates while playing the replay? If yes then I'd reccomend calculating them first, saving them into an array and the use the saved values.
Signature temporarily out of order.
Originally Posted by psycore
The body rotation just tells you how the bodypart is rotated around its center, not around its joint.
but it should be equal to joint rotation, isn't it?
Originally Posted by psycore
If what you wrote in 2. already has to much operations I don't see much you can do. I'm not sure how the "^" operator works but it may be faster if you just write x*x instead of x^2.
I don't quite get which angle you're trying to get. I made a little sketch and it doesn't seem to make any sense:
I assume that you want the angle between biceps and triceps to measure how far the elbow is rotated. To get that you'll need far more operations than those you wrote.
Yes, it should be calculated for every axis. The problem is it's 5*3*40 operations per frame. And yeah, that should some more 'magic' operations, that would make only one axis rotation from three of them. I'm not sure how it works.
Originally Posted by psycore
Are you calculating the coordinates while playing the replay? If yes then I'd reccomend calculating them first, saving them into an array and the use the saved values.
No, during replay i only save world postitions of joints. Even with it, i got some frame-skipping problems. Data is ok, because it freezes every frame, but i dont know, how to remove gap between freeze and unfreeze, cuz threre's no "shift+space" command.
HAMPA SUX!
URL removed ~SF
URL removed ~SF
Sorry, took me a while to figure some stuff out.
It's possible that we're not speaking of the same thing. With joint rotation I meant the angle between the two bodyparts conected by the joint. If that's what you meant, too, then joint rotation is not equal to body rotation. Imagine you hold your elbow but raise your shoulder. The angle between biceps and tricepts (the joint angle of the elbow) did not change but the triceps rotated.
But if you wanted to know how far you have to rotate the original (unrotated) bodypart to get the current rotation, this might help:
"Lua code:
At least this is what I got after some time of thinking. I checked it with the draw_box() command and it seems that it has the wrong sign. I don't know if I made a mistake in my calculations or draw_box() rotates objects the wrong way round. Just check it.
run_frames(1) is like shift+space.
Originally Posted by rfxDarth
but it should be equal to joint rotation, isn't it?
It's possible that we're not speaking of the same thing. With joint rotation I meant the angle between the two bodyparts conected by the joint. If that's what you meant, too, then joint rotation is not equal to body rotation. Imagine you hold your elbow but raise your shoulder. The angle between biceps and tricepts (the joint angle of the elbow) did not change but the triceps rotated.
But if you wanted to know how far you have to rotate the original (unrotated) bodypart to get the current rotation, this might help:
"Lua code:
local bpRot = get_body_info(player, bp).rot
local x = math.atan2(-bpRot.r6, bpRot.r10)
local y = math.asin(bpRot.r2)
local z = math.atan2(-bpRot.r1, bpRot.r0)
At least this is what I got after some time of thinking. I checked it with the draw_box() command and it seems that it has the wrong sign. I don't know if I made a mistake in my calculations or draw_box() rotates objects the wrong way round. Just check it.
Originally Posted by rfxDarth
cuz threre's no "shift+space" command.
run_frames(1) is like shift+space.
Signature temporarily out of order.
