Support NeoGAF

[gofreak]

That sound more like a limitation of the physics engine than the controls itself though.

Whatever about the physics, if you can't in the first place move the paddle up and in to the net where you want it to meet the ball, you can't make that shot.

 

[dragonfart28]

That sound more like a limitation of the physics engine than the controls itself though.

Yes, that isn't something Motion+ wouldn't be able to do. Grand slam tennis is a pretty good example of this sort of thing and is entirely possible with that game.

 

[Earl Cazone]

itas also really fucking impossible to make a nice sideways spinning dropshot right behind the net, that jumps of close to 90 degrees behind.
its my favourite shot in regular table tennis and i cant get it done in SC

 

[Lonely1]

Everything drifts. Its about how you recalibrate the system to error correct.

The camera in move recenters absolute X,Y,Z position when it sees the ball this is what the Wiimote actually cannot do well.

The Wiimote however does have a camera pointing at the sensor bar. So it is visually attempting to figure out what you are pointing at.

Move is technically "blind" to what you are pointing at. It already has absolute X,Y,Z position so it must rely on gyroscope pointing which with enough resolution in the gyroscope should work fine.

Ideally, the Wiimote would have a sensor bar light at all 4 corners of the TV. Then it would have much better absolute pointing. As it is right now if you have a 70 inch TV the Wii doesn't handle that properly and its a well known issue.

The larger the TV ironically, the better the Gyroscope pointing will perform in Move.

Well, that's is true when trying to emulate RL aiming. But when people, at least me, talk about pointing like a 3D mouse we talk more about the abstraction of movement used to effectively move a cursor around the screen, like a mouse doesn't do a 1-1 map of screen to the mouse-pad. And this aspect is what (imo) is more important for core gaming.

 

[distrbnce]

That sound more like a limitation of the physics engine than the controls itself though.

How is it a limitation that the physics system can detect when you are blocking the path of the ball (rather than when you waggle to signify a hit)?

Is this a limitation of our universe's physics as well? This is how many returns are done in ping pong.

 

[gofreak]

How is it a limitation that the physics system can detect when you are blocking the path of the ball (rather than when you waggle to signify a hit)?

Is this a limitation of our universe's physics as well? This is how many returns are done in ping pong.

I think he's arguing that it's an issue of the physics engine, not whether there's positional tracking.

My reply is that regardless of the physics, you need positional tracking for that.

Yes, that isn't something Motion+ wouldn't be able to do. Grand slam tennis is a pretty good example of this sort of thing and is entirely possible with that game.

GST isn't remotely doing the same thing.

But to abstract for a moment and come back to your original argument, forgetting about specific titles, your argument was that there's no real difference gameplay wise between a table tennis game that uses positional tracking and one that doesn't. That's a claim that just doesn't stand up.

 

[blurredvision]

I only read through the first couple hundred posts in here, so my apologies if this has been asked. My personal opinion is that the tech behind Move is fantastic, but unfortunately I've never used WM+ or played Wii Sports Resort (only the original launch controller and Wii Sports).

The common comparison between Move and WM+ is table tennis, which is a legitimate comparison given the full range of motion needed in all aspects. However, it's hard to compare IMO given that in WSR, you have no control over your character's movement. If the ball is hit softly in front of the net, your Mii is moved up automatically and you hit it. If it's hit towards the back of your person, your Mii is moved back automatically and you hit it. In Sports Champions, your in-game persona (i.e. floating paddle) is static while it's up to you, with controller in hand, to actually step forward and back in your living room in order to hit short or deep balls.

Now, here comes my question. In regards to Move tracking in 3D space and requiring the user to respect it, it's awesome. Is the WM+ capable of mimicking the Move in this regard, assuming the software was written for it? In other words, if we just took the WM+ tech and moved it over as-is to the PS3, could you fully play Table Tennis in Sports Champions without any modifications needed to the tech or the game?

I'm curious on this, since I've not had any hands-on time with the WM+.

 

[Lonely1]

I think he's arguing that it's an issue of the physics engine, not whether there's positional tracking.

My reply is that regardless of the physics, you need positional tracking for that.

Well, is true that WM+ won't allow for the range of movements needed to move the racket in position in the way you're describing, so the avatar needs to move themselves. But the fact that you can't merely block the balls trajectory and need to do the swing motion to return every ball is a limitation of the physics engine. Just like you need to press a button in order to block in WSR Sword game and simply positioning the sword in the right angle won't do.

 

[UntoldDreams]

These games arguments are somewhat "fuzzy" at best but there is a simple point here to be made.

Wii games are designed with Wii hardware in mind. By definition, you don't need the extra capabilities of MOVE to play any Wii games.

The opposite though is a problem.

You cannot put absolute positioning Table Tennis Sports Champions on the Wii. You simply can't without changing the game back into a WSR type of game.

 

[poppabk]

Now, here comes my question. In regards to Move tracking in 3D space and requiring the user to respect it, it's awesome. Is the WM+ capable of mimicking the Move in this regard, assuming the software was written for it? In other words, if we just took the WM+ tech and moved it over as-is to the PS3, could you fully play Table Tennis in Sports Champions without any modifications needed to the tech or the game?

I'm curious on this, since I've not had any hands-on time with the WM+.

It has no really reliable way of measuring its location in 3d space, so no, not with the same level of accuracy (and more importantly reliability) for translational movements. Add a camera and it could, even with just eyetoy type body tracking.

 

[UntoldDreams]

I only read through the first couple hundred posts in here, so my apologies if this has been asked. My personal opinion is that the tech behind Move is fantastic, but unfortunately I've never used WM+ or played Wii Sports Resort (only the original launch controller and Wii Sports).

The common comparison between Move and WM+ is table tennis, which is a legitimate comparison given the full range of motion needed in all aspects. However, it's hard to compare IMO given that in WSR, you have no control over your character's movement. If the ball is hit softly in front of the net, your Mii is moved up automatically and you hit it. If it's hit towards the back of your person, your Mii is moved back automatically and you hit it. In Sports Champions, your in-game persona (i.e. floating paddle) is static while it's up to you, with controller in hand, to actually step forward and back in your living room in order to hit short or deep balls.

Now, here comes my question. In regards to Move tracking in 3D space and requiring the user to respect it, it's awesome. Is the WM+ capable of mimicking the Move in this regard, assuming the software was written for it? In other words, if we just took the WM+ tech and moved it over as-is to the PS3, could you fully play Table Tennis in Sports Champions without any modifications needed to the tech or the game?

I'm curious on this, since I've not had any hands-on time with the WM+.

As I said Wii games are made for the Wii. The Move can basically mimic Wiimote functionality at an acceptable or superior level so porting should work from Wii to PS3 as far as game design.

Its going the other way which would be an issue if the game was designed with the specific intention of Move 3D spatial positioning.

 

[Vinci]

These games arguments are somewhat "fuzzy" at best but there is a simple point here to be made.

Wii games are designed with Wii hardware in mind. By definition, you don't need the extra capabilities of MOVE to play any Wii games.

The opposite though is a problem.

You cannot put absolute positioning Table Tennis Sports Champions on the Wii. You simply can't without changing the game back into a WSR type of game.

This is actually one of the best arguments I've seen thus far about the distinction between these two products. For one thing, it lacks bias of any kind. Good post.

 

[Jeels]

Now, here comes my question. In regards to Move tracking in 3D space and requiring the user to respect it, it's awesome. Is the WM+ capable of mimicking the Move in this regard, assuming the software was written for it? In other words, if we just took the WM+ tech and moved it over as-is to the PS3, could you fully play Table Tennis in Sports Champions without any modifications needed to the tech or the game?

I don't think Wii Motion + can calculate your absolute position in space. Maybe someone can correct me if I'm wrong.

 

[Koren]

Now, here comes my question. In regards to Move tracking in 3D space and requiring the user to respect it, it's awesome. Is the WM+ capable of mimicking the Move in this regard, assuming the software was written for it? In other words, if we just took the WM+ tech and moved it over as-is to the PS3, could you fully play Table Tennis in Sports Champions without any modifications needed to the tech or the game?

Short answer: no.

Accelerometers and gyroscopes, theorically, should give you full 3D position tracking by integration, assuming that you know the start point. Again, that's in theory. In practice, the computation are really complex, rounding errors have big consequences, and anytime you reach the maximum capability for accelerometers/gyroscopes, you'll lose 3D position/angle. Submarines do this, planes used to do this before GPS, but Wii tech isn't precise enough.

You can do motion sensing (computing parameters of a ball strike, of a swing) but forget about doing any complete and reliable 3D motion tracking.

 

[Jokeropia]

Not exactly.

AAA AAA

The above A's represent what the wiimote camera sees looking at the sensor bar lights.
That is actually not enough to extrapolate your absolute position in 3D space since it has no real way to determine your position.

As it knows the actual distance between the lights on the sensor bar and how far apart they appear to be, it knows the absolute distance. This is all described very throughly in the patent application. Here's an extract:

[0123]At step 53, the CPU 30 performs the distance calculation process, and then proceeds to the next step. In the distance calculation process, the current distance realD between the controller 7 and the markers 8L and 8R is calculated based on the first coordinate data Da1 and second coordinate data Da2, which have been transmitted from the controller 7 and stored in the main memory 33. Hereinafter, processes performed at step 53 will be described in detail with reference to FIGS. 16 and 17.

[0124]As shown in FIG. 16, the CPU 30 obtains the first coordinate data Da1 and second coordinate data Da2 (step 71), and calculates a distance mi (step 72). As shown in FIG. 17, the distance mi is a distance between two points in the taken image. These two points correspond to images of the markers 8L and 8R in the taken image, and coordinates indicating the two points are the first coordinate data Da1 and second coordinate data Da2. Accordingly, the CPU 30 uses the first coordinate data Da1 and second coordinate data Da2 to calculate the distance mi. To be specific, when the first coordinate data Da1 is position coordinates (Lx, Ly) and the second coordinate data Da2 is position coordinates (Rx, Ry), the distance mi is obtained by the following equation.

mi = ( Rx - Lx ) 2 + ( Ry - Ly ) 2 [ equation 1 ]

[0125]Next, the CPU 30 calculates a width w (refer to FIG. 17) which indicates, with respect to setting positions of the markers 8L and 8R, a width for which the image pickup element 743 is able to take an image (step 73). The width w is obtained by the following equation.

w=wi.times.m/mi

Here, m represents a setting distance between the markers 8L and 8R (actual setting distance between the the markers 8L and 8R; e.g., 20 cm), and is a fixed value. Also, wi represents a width wi of the image taken by the image pickup element 743 which corresponds to the width w. The width wi is also a fixed value. Since the setting distance m and width wi are fixed values, these values are prestored in storage means (not shown) within the game apparatus 3. Note that, the player is allowed to discretionarily determine the setting positions of the markers 8L and 8R in accordance with the player's environment, thereby determining the setting distance m. In such a case, the player is required to input a distance between the discretionarily determined setting positions of the markers 8L and 8R as the setting distance m so that the width w can be obtained from the above equation.

[0126]Next, the CPU 30 calculates the current distance realD (refer to FIG. 17) between the image pickup element 743 (controller 7) and the markers 8L and 8R, by using the width w and a viewing angle .theta. of the image pickup element 74, and updates the current distance data Db3 (step 74). Then, the distance calculation process in the subroutine ends. Here, the current distance realD is obtained by using the following equation.

realD=(w/2)/{tan(.theta./2)}

Since the viewing angle .theta. is a fixed angle, the angle .theta. is prestored in the storage means (not shown) within the game apparatus 3.

It also calculates the distances realDL and realDR which is the distance from the IR camera and the left and right marker respectively:

[0149]Next, by using the width w calculated at step 102 and a viewing angle .theta. of the image pickup element 743, the CPU 30 calculates a current distance realDL (see FIG. 22) which is a current distance between the marker 8L and the image pickup element 743 (controller 7) (step 103). The current distance realDL is obtained from the following equation.

realDL=(w/2)/{tan(.theta./2)}

Since the viewing angle .theta. is a fixed angle, the viewing angle .theta. is prestored in the storage means (not shown) within the game apparatus 3.

[0150]Next, based on the second size data (diameter diamR), the CPU 30 calculates another width w which indicates, with respect to a setting position of the marker 8R, a width for which the image pickup element 743 is able to take an image (step 104). Said another width w is obtained from the following equation.

w=wi.times.diamM/diamR

[0151]Then, by using said another width w calculated at step 104 and the viewing angle .theta. of the image pickup element 743, the CPU 30 calculates a current distance realDR between the marker 8R and the image pickup element 743 (controller 7) (step 105). Here, the current distance realDR is obtained from the following equation.

realDR=(w/2)/{tan(.theta./2)}

Using these values it can then calculate the angular position:

[0154]It is also possible to obtain an angle .delta.2 between a line connecting the controller 7 and said middle point and the line connecting the markers 8L and 8R. The angle .delta.2 is obtained from the following equation based on the cosine theorem by using the current distance realD, current distance realDL and setting distance m.

cos .delta.2={realD.sup.2.times.(m/2).sup.2-realDL.sup.2}/{2.times.realD.t- imes.(m/2)}

Obtaining the angle .delta.2 allows the CPU 30 to calculate an angular position of the controller 7 with respect to the markers 8L and 8R, and various processes can be performed in accordance with the angular position of the controller 7.

Further, as distance increases the ability to calculate your position becomes significantly worse due to the resolution of the Wiimote camera.

The Wii uses 8x subpixel analysis to provide 1024x768 resolution for the tracked points, so this is not an issue during normal use.

Position and orientation (x, y, z, r, t, w) : 6 values

position of two spots in the camera space (u1, v1, u2, v2) : 4 values

No it's not sufficient. rotation of the wiimote around the axis between the two leds (raise and look down) and the rotation on a circle that pass through both led can't be tracked with camera.

Are you referring to how holding the Wiimote, say, 30 cm directly above the sensor bar and pointing down should look the same as holding it 30 cm below and pointing up? In these cases the orientational data from the gyroscope in MotionPlus can help supply the direction.

 

[Windom Earle]

As I mentioned earlier, the wiimote seems to detect any significant IR as the sensor bar. There is an option in WSR that allows the motion+ to recalibrate off the sensor bar on the fly when it sees it, however, it can screw up if it sees an alternate source (this can happen a lot in archery where your remote is pointing directly up usually at a light fixture). If you turn this feature off, then you see a steady drift from the gyros. Recalibrating is as easy as pointing at the screen and pressing down, but having to do it manually is annoying.

Thank you, I will investigate (Even though I don't play it very often)

 

[UntoldDreams]

As it knows the actual distance between the lights on the sensor bar and how far apart they appear to be, it knows the absolute distance. This is all described very throughly in the patent application. Here's an extract:
It also calculates the distances realDL and realDR which is the distance from the IR camera and the left and right marker respectively:

Using these values it can then calculate the angular position:

The Wii uses 8x subpixel analysis to provide 1024x768 resolution for the tracked points, so this is not an issue during normal use.
Are you referring to how holding the Wiimote, say, 30 cm directly above the sensor bar and pointing down should look the same as holding it 30 cm below and pointing up? In these cases the orientational data from the gyroscope in MotionPlus can help supply the direction.

Arguing the actual patent application is not going to help.

However... To answer your points. Yes roughly it works fine.
The problem is as I stated before: distance and resolution.

TO CLARIFY: I can tell you how far you are away from me very accurately when you are standing up to 5 feet away from me. I might even be able to tell an inch movement left right forward backward accurately.

However, if you are 10 feet away from me I will be absolutely unable to tell if you drift one inch left right forward back.

That is the point here. The patent is fine. The math is fine. The resolution and cost of the wiimote camera versus the distance and speed of data acquisition is the problem.

This is all beyond that patent description my friend.

Again... The issue is that without a super high resolution camera you simply cannot make the system work properly at any reasonable distance because you need SUB-PIXEL detail to determine if I shift one inch to the left or right.

Keep in mind sir that I think Nintendo was brilliant here. Its nothing to be ashamed of that your competitor took 3+ years to get something better.

Also, keep in mind the reason you don't see awesome Nintendo created spatial positioning games all over the place is because they already KNOW about this limitation and actively design games (which are fun) which do NOT rely on spatial positioning.

 

[distrbnce]

As it knows the actual distance between the lights on the sensor bar and how far apart they appear to be, it knows the absolute distance. This is all described very throughly in the patent application. Here's an extract:
It also calculates the distances realDL and realDR which is the distance from the IR camera and the left and right marker respectively:

Using these values it can then calculate the angular position:

The Wii uses 8x subpixel analysis to provide 1024x768 resolution for the tracked points, so this is not an issue during normal use.
Are you referring to how holding the Wiimote, say, 30 cm directly above the sensor bar and pointing down should look the same as holding it 30 cm below and pointing up? In these cases the orientational data from the gyroscope in MotionPlus can help supply the direction.

It's a strange day on gaf when to explain what the Wii-mote can do, crazy technical specs are brought out and lauded. Like Sony's 2006 all over again.

Obviously, these technical specs are pretty dense, so what's the best real-world example of what Wii can do in 3D space? Are there other examples of "realD" in practice?

 

[Jokeropia]

Again... The issue is that without a super high resolution camera you simply cannot make the system work properly at any reasonable distance because you need SUB-PIXEL detail to determine if I shift one inch to the left or right.

It's a good thing it uses subpixel analysis for this, then.

 

[Koren]

Are you referring to how holding the Wiimote, say, 30 cm directly above the sensor bar and pointing down should look the same as holding it 30 cm below and pointing up? In these cases the orientational data from the gyroscope in MotionPlus can help supply the direction.

Help, but:
- it's second order
- there's difficulties to separate motion and position from accelerometers readings
etc.

But the second one is more difficult to get even with accelerometers and gyroscopes. Imagine a circle that go through bot leds :

You see the same at both points...

 

[cakefoo]

Has anyone talked about AiLive?

Remember the video of the engineer doing the pseudo 1:1 gestures?

Every frame of this gif represents the first frame after resetting the scene for his next gesture. Notice how his controller position changes from reset to reset, but the virtual object position keeps spawning right I drew that white dot. That's because's recalibrating for the start/home position so it'll have the best chance to track his gesture with convincing enough accuracy as is possible with those accelerometers.

The only way Wii could do Move table tennis is if the player recalibrates before every lateral input, which would just be insanely impractical.

Wii's positional tracking is one of those "close but no cigar" deals.

 

[UntoldDreams]

It's a good thing it uses subpixel analysis for this, then.

I think... We have a gap between my mathematical and engineering capabilities and yours.

I do not wish to insult you but what you are saying is illogical. You seem like a decent person so I hope you can come around and see the light.

By saying "SubPixel Data" is required to do this calculation... It means by definition that the image data you currently have is not acceptable to perform the task.

When people create subpixel data they are attempting to extrapolate something between two or more neighboring pixels.

In this case, the Wiimote Camera would have multiple images which lack the resolution to be different enough to calculate position.

Let me put it this way. Let's say the resolution of the Wiimote Money Scanner can tell me if you (rounded) have $10 or $20 or $30 or some multiple of 10.

The Wiimote Money Scanner tells me you have $30. The problem here is that is not high enough resolution. I need to know EXACTLY if you have $29 or if you have $31.

Simply saying $30 is not enough RESOLUTION for me to determine your absolute position.

That is the limitation of the Wiimote camera I hope you can understand. Its simply not detail enough to give accurate 1:1 spatial positioning.

 

[Branduil]

So, what would be the best way to incorporate the best feature of both WM+ and Move in a future controller, so you have the pointing accuracy of WM+ and the positional accuracy of Move?

 

[poppabk]

The resolution and cost of the wiimote camera versus the distance and speed of data acquisition is the problem.

Also, keep in mind the reason you don't see awesome Nintendo created spatial positioning games all over the place is because they already KNOW about this limitation and actively design games (which are fun) which do NOT rely on spatial positioning.

The camera has sub-pixel resolution of 1024x768 at 120hz (I think). The limitation is that the remote has to be pointed at the camera, which will only ever work for a tiny subset of games.

 

[cakefoo]

P.S. I don't know how much more credibility an actual Wiimote homebrew programmer has over someone with the name "Jokeropia" but, according to him you would need a 4-point, tetrahedron-shaped IR beacon to determine XYZ.

http://www.youtube.com/watch?v=KyvIlKSA0BA

 

[poppabk]

I think... We have a gap between my mathematical and engineering capabilities and yours.

I do not wish to insult you but what you are saying is illogical. You seem like a decent person so I hope you can come around and see the light.

By saying "SubPixel Data" is required to do this calculation... It means by definition that the image data you currently have is not acceptable to perform the task.

Both move and the wii remote are using sub-pixel resolution. Otherwise the 320x240 (only a small portion of which would be used) of the PS3 camera when extrapolated to a big screen would result in woeful resolution of movements of the balls.

 

[Jomjom]

So, what would be the best way to incorporate the best feature of both WM+ and Move in a future controller, so you have the pointing accuracy of WM+ and the positional accuracy of Move?

Wait it's fact that WM+ has better pointing accuracy than the Move? I've had WM+ since launch and if Move has worse pointing accuracy than that, I think I'll just skip out on Move completely. The WM+ was hugely disappointing to me as it was not at all accurate like Nintendo marketed it to be.

 

[poppabk]

Wait it's fact that WM+ has better pointing accuracy than the Move? I've had WM+ since launch and if Move has worse pointing accuracy than that, I think I'll just skip out on Move completely. The WM+ was hugely disappointing to me as it was not at all accurate like Nintendo marketed it to be.

The regular wiimote has great pointing accuracy, M+ has nothing to do with it. What do you mean by inaccurate pointing?

 

[distrbnce]

Wait it's fact that WM+ has better pointing accuracy than the Move? I've had WM+ since launch and if Move has worse pointing accuracy than that, I think I'll just skip out on Move completely. The WM+ was hugely disappointing to me as it was not at all accurate like Nintendo marketed it to be.

Not so much fact, but rather a strange debate some users here have grasped onto as potentially something that's equal between the two devices.

I (being someone that owns Move) think it's as accurate as it possibly could be in RE5, but I'm clearly not as technically minded as these other forum-goers.

 

[defferoo]

Has anyone talked about AiLive?

Remember the video of the engineer doing the pseudo 1:1 gestures?

Every frame of this gif represents the first frame after resetting the scene for his next gesture. Notice how his controller position changes from reset to reset, but the virtual object position keeps spawning right I drew that white dot. That's because's recalibrating for the start/home position so it'll have the best chance to track his gesture with convincing enough accuracy as is possible with those accelerometers.

The only way Wii could do Move table tennis is if the player recalibrates before every lateral input, which would just be insanely impractical.

Wii's positional tracking is one of those "close but no cigar" deals.

Those weren't pseudo 1:1 gestures, sure they may lose some accuracy if you swing too hard, but they did correctly map player movements to objects on screen. see http://www.youtube.com/watch?v=acND4sO3pJs at 0:50. If that is pseudo 1:1, then Move should also be pseudo 1:1. And yes, you have to recalibrate, but the main difference is that the Move is constantly recalibrating with the digital compass and the camera, and the Wii only gets a chance to calibrate when the pointer happens to see the sensor bar or when the remote isn't moving.

I know for a fact that the Move's various sensors are more sensitive than the Wii/Motion+ sensors. The gyro can cover greater degrees of rotation/second and the accelerometer might be able to handle more Gs. But the real difference between the two lies in the camera implementation and the fact that Move has a compass.

The Wii remote can detect the relative position of where the remote is pointing more accurately than the Move, but the Move is better at detecting the relative position of where the Move controller is. The way I see it, the Move controller's "pointing" uses a combination of the camera, accelerometer, and gyroscope (possibly compass too). By moving the Move, it will calculate the change in angle and use that to change the cursor location, similar to how red steel 2 uses motion+ to allow you to keep turning when your cursor is off screen. However, this isn't as precise as the Wii remote's pointer which doesn't need to use the accelerometer and gyroscope at all.

This makes the Move better at relative position tracking aka 1:1 motion tracking, and the Wii remote better at relative pointing. Now... if they added a camera to the sensor bar and lights and a compass to the Wii remote... haha

 

[UntoldDreams]

The camera has sub-pixel resolution of 1024x768 at 120hz (I think). The limitation is that the remote has to be pointed at the camera, which will only ever work for a tiny subset of games.

...
Let's clarify a definition first.

A camera does not have "subpixel" resolution at the digitizing level. A camera has a hardware limit on its actual PHYSICAL PIXEL data based on its lens & digitizing element.

The camera internally, after taking a snapshot picture... it can run a math algorithm to try and create subpixel data from the original camera snapshot creating an anti aliasing effect and smoothing out an image.

Let's say these are 5 pixels:
XXXXX

At close range these 5 pixels are highly representative of about 1 inch of spatial data.
Meaning my finger close to the camera can be accurately detailed by these 5 pixels.

Now let me move my finger 10 feet away.

At this point. My finger is so small its exactly 1 pixel in size.
00X00

See... That X represents my finger seen by a camera 10 feet away.

Now let me move my finger 1 inch forward, left, right, backward.
00X00

The image did not change in the 5 pixels! The problem is that the resolution of the camera is not HIGH ENOUGH to capture the detail of my tiny finger moving at 10 feet one inch left or right or forward or backward.

In fact this same issue will plague Kinect and why they have to give up on finger tracking.


You cannot create this type of important data with an algorithm. It must be captured in the first place.

 

[UntoldDreams]

Both move and the wii remote are using sub-pixel resolution. Otherwise the 320x240 (only a small portion of which would be used) of the PS3 camera when extrapolated to a big screen would result in woeful resolution of movements of the balls.

I think I explained what subpixel means above.

SubPixel data is a calcuation from raw data. It does not magically create more information.

If your original data is not high enough there is nothing you can do.

As to your point about MOVE and the PS Eye its not the same.

The Wiimote is attempting to figure out its OWN position based on two tiny points of light very far away. That is actually very HARD.

The PSEye is a stationary camera which is attempting to figure out your line of sight point position. It actually doesn't need much more than the CENTER OF THE GLOWING BALL AS A POINT REFERENCE to mathematically calculation your position.\

EDIT:
Move will have the same problem at a certain distance by the way.
At some point the move ball is a single pixel on the camera. And moving one inch left and right will not produce a different image for the Eye. At that point there is no longer any error correction from the camera at that distance. Once you move enough though it will continue to recalibrate because it can see the "pixel point size ball" moving.

 

[Jokeropia]

But the second one is more difficult to get even with accelerometers and gyroscopes. Imagine a circle that go through bot leds

Right, that image looks like what I referred to. In position A, the controller is pointing down while in position B, it is pointing up.

Edit: Wait, I might've misunderstood. We're looking at the image from above, yes? And to get from point A to point B you don't move the controller down, you move it sideways?

If so, you don't actually need the gyroscope to tell your position (though it will tell the difference between pointing left and pointing right as well), since the Wiimote calculates the distance from each markers separately. (In the patent application, these distances are referred to as realDL and realDR respectively.) At point A, the distance to the right marker is greater than the distance to the left marker, but at point B it's the other way around.

I think... We have a gap between my mathematical and engineering capabilities and yours.

:lol

By saying "SubPixel Data" is required to do this calculation... It means by definition that the image data you currently have is not acceptable to perform the task.

This is from the WiiBrew wiki:

The built-in processor uses 8x subpixel analysis to provide 1024x768 resolution for the tracked points.

Are you talking about subpixels in a different context?

Obviously there's a limit to everything as you can break down distances into infinitely small units, but if I took a 1024x768 picture of you from across the living room, then moved a bit and took another picture, I'd have to have moved a really short distance for the pictures to come out pixel-identical.

 

[mrklaw]

Untolddreams, question for you.

Assuming the user has calibrated by pointing at the top left and bottom right of the screen, should move and wiimote be similarly capable of aiming at a point on the screen accurately like a light gun?

I've seen wii games do this well, but not move yet (obviously it's only just cone out). The Shoot is good, but I found it losing accuracy when I did some of the other actions, eg turning the move to the left and pumping up and down to move the train. When pointing back at the screen it'd often be registering to the left of where I'm pointing.

 

[UntoldDreams]

Right, that image looks like what I referred to. In position A, the controller is pointing down while in position B, it is pointing up.
:lol
This is from the WiiBrew wiki:
Are you talking about subpixels in a different context?

Obviously there's a limit to everything as you can break down distances into infinitely small units, but if I took a 1024x768 picture of you from across the living room, then moved a bit and took another picture, I'd have to have moved a really short distance for the pictures to come out pixel-identical.

Good lord.

*** What you are seeing:
UntoldDreams is babbling. I just have to find the right wiki link to beat him up.

*** What I am seeing:
A person I am trying to educate on this topic. Since I am educated on this topic as its been part of my job when I worked in Vision Processing Systems I feel like... I should be able to gently educate this guy into being a better person. No need to come in and stomp on the guy.

 

[cakefoo]

Those weren't pseudo 1:1 gestures, sure they may lose some accuracy if you swing too hard, but they did correctly map player movements to objects on screen. see http://www.youtube.com/watch?v=acND4sO3pJs at 0:50. If that is pseudo 1:1, then Move should also be pseudo 1:1.

Uhh... NO? Move only goes 1/60th of a second between recalibrations, meaning that any drifting is corrected WELL before it becomes noticeable to the human eye. Wii has to go much longer between checks. So long that massive drifting will occur. Also the video you see there is hard to gauge how 1:1 the accelerometers even are. You can't see his hand in the full motion and you don't get to see what happens when you rely on the accelerometers alone for more than one slow swipe, so that video is not going to alleviate any concerns.

Besides, if you read the pdf manual on their site, you'll see that they confess that it's not good enough for more than one gesture at a time due to accelerometer drift--- kind of hypocritical considering they were just claiming "full 1:1 3D tracking" a moment prior.

And yes, you have to recalibrate, but the main difference is that the Move is constantly recalibrating with the digital compass and the camera, and the Wii only gets a chance to calibrate when the pointer happens to see the sensor bar or when the remote isn't moving.

Yes but that's exactly the problem. With time the accelerometer's positional estimates decrease in accuracy so much that they become dislodged from the true 3D position, and this would frustrate a player incredibly, so devs don't even TRY it with the Wii except for slow, calm movements like in WSR archery.

 

[Graphics Horse]

If your original data is not high enough there is nothing you can do..

Yes, but you're thinking of the input array as being a binary image with a single ray of light for each camera cell, what you'll really get is fractions of finger/sensor light per cell when it's between cells, instead of just on and off. This is where you get the sub pixel accuracy from.
If your finger is close enough to appear on video, it will always be visible and move smoothly from cell to cell. doesn't flicker on and off when you move your hand.

 

[cakefoo]

Right, that image looks like what I referred to. In position A, the controller is pointing down while in position B, it is pointing up..

that's a top-down diagram, dude. The pink is the sensor bar and A and B are where you're standing off-center left and off to the right. The Wiimote can't tell the difference between A and B due to the 2 points on the sensor bar looking the exact same from both positions. That's why I believe this Wii homebrewer when he says a 4-point sensor bar in the shape of a tetrahedron is needed.

http://www.youtube.com/watch?v=KyvIlKSA0BA

 

[Jomjom]

The regular wiimote has great pointing accuracy, M+ has nothing to do with it. What do you mean by inaccurate pointing?

The pointer jitters, it doesn't point exactly where I am pointing, and it lags.

 

[donny2112]

gently educate this guy into being a better person.

It's probably better to not talk down to people when you're trying to "educate" them.

 

[UntoldDreams]

Yes, but you're thinking of the input array as being a binary image with a single ray of light for each camera cell, what you'll really get is fractions of finger/sensor light per cell when it's between cells, instead of just on and off. This is where you get the sub pixel accuracy from.
If your finger is close enough to appear on video, it will always be visible, doesn't flicker on and off when you move your hand.

Yes more accurate extrapolation can be inferred from higher resolution internal data beyond the hard pixels.

That being said...the point is that the fuzzy factor is too great for what people are asking from the Wiimote camera.

A-----------------------------------------BB

This distance is you have to run an algorithm against and error correct with.
Its not just hard its quite plainly inaccurate past a certain distance from the wiimote (A) to the sensor bar (BB).
EDIT: Again this effects any camera system including MOVE and KINECT

 

[UntoldDreams]

It's probably better to not talk down to people when you're trying to "educate" them.

I apologize if it comes out condescending but to be clear, the point is not "I am super" its that people should be open to listening sometimes.

If you are wrong accept it and move on. Its how you have a better life.

Plowing through a bunch of wikis in an attempt to argue with someone who is patiently trying to explain things is... counterproductive.

 

[Graphics Horse]

Its not just hard its quite plainly inaccurate past a certain distance from the wiimote (A) to the sensor bar (BB).

You're right, but it's fine from a distance people play from.

 

[poppabk]

...
Let's clarify a definition first.

A camera does not have "subpixel" resolution at the digitizing level. A camera has a hardware limit on its actual PHYSICAL PIXEL data based on its lens & digitizing element.

But anything which is moving will create aliasing which allows its position to be calculated more accurately based on interpolation. Sure there is a distance where you will see nothing but a single pixel point of light and you are no longer able to interpolate anything, but you are talking beyond living room distances for this.

The actual resolution of the wiimote camera is 128x96 and it has a field of view of approximately 30 degrees by 20 degrees. For a normal viewing distance (I used the THX recommended viewing distance for a 42" TV) this results in the sensor bar being in view over an x distance of about 75cm (which is the maximum you can move laterally before you lose sight of the sensor bar). So a pixel equates to approximately 6mm. Each of the two LED light patterns on the sensor bar is probably 3cm wide, therefore without interpolation you can still resolve individual points, allowing for sub-pixel interpolation to occur. Using the 1024 sub pixel accuracy - each interpolated pixel will equal 7/10ths of a mm.

 

[UntoldDreams]

You're right, but it's fine from a distance people play from.

I think the part of the discussion you aren't involved with was when people were wondering why the Wii doesn't have games like One to One motion in Sport's Champions ping pong.

The short general response was that its "hard" for the Wii to have 1:1 positional accuracy during gameplay.

I was attempting to explain WHY its hard so people have a point of reference.

Instead, we went down a long arduous path where people said "No the Wii can do it just as good" despite the reality of... Not having a bunch of games which are 1:1 positional like Sports Champions.

Again... I think Nintendo is brilliant. They won the race and they deserve it.

 

[jiggles]

I'll not have anyone trash the integrated image processing chip on the Wiimote. No sir. As someone who has studied computer vision, and worked extensively with the hardware, the accuracy of the data they manage to extrapolate from a 128x96 image is fucking godly.

 

[poppabk]

I think the part of the discussion you aren't involved with was when people were wondering why the Wii doesn't have games like One to One motion in Sport's Champions ping pong.

The short general response was that its "hard" for the Wii to have 1:1 positional accuracy during gameplay.

I was attempting to explain WHY its hard so people have a point of reference.

But your explanation has nothing to do with that. The reason that the Wii is incapable of doing that is because the camera can't see the sensor bar at all times because it is on the end of the remote, the move camera can see the ball on the remote at all times because the camera is stationary and the ball is moving within the cameras view. The effective resolution of the move camera is most likely lower than the wiimote camera, as its field of view is much larger. Both use sub-pixel interpolation to make up for some of the defects in their physical resolution.

 

[donny2112]

I apologize if it comes out condescending

You told him there was a gap (implying in your favor) between his math/engineering skills and yours. Even if that's the case, pointing it out is a weak argument, as it's basically a "I'm better than you" reasoning.

Plowing through a bunch of wikis in an attempt to argue with someone who is patiently trying to explain things is... counterproductive.

As long as the person is seeking to find the correct answer, it's not. If they're just trying to gain a clearer understanding and something you say doesn't match up with a trusted source of theirs, it's fine to pull in other sources to either prove or disprove a position (e.g. the Wiki's point or yours). If they're doing it to just be argumentative, then, yeah, it's not going to amount to much.

 

[Graphics Horse]

I think the part of the discussion you aren't involved with was when people were wondering why the Wii doesn't have games like One to One motion in Sport's Champions ping pong.

Alright, I think you were correct in saying it's not accurate enough to judge exact spatial position of the Wiimote from the lights, I was just objecting to your explanation of subpixel differences as being undetectable.

 

[distrbnce]

Obviously, these technical specs are pretty dense, so what's the best real-world example of what Wii can do in 3D space? Are there other examples of "realD" in practice?

Anyone? Can't we get two discussions going on here? The boring, theoretical possibilities of both systems and the actual real-world experiences that can be had right now?