Mechanical exoskeleton: a New hand motion capturing approach
Hey Guys, my name is Aler, me and my team(Dexta Robotics @ "www.dextarobotics.com") have been working on a new type of hand motion capturing approach since last September. Over the year we made a lot of progress on making a truly affordable, easy to use and precise hand motion capturing device, called Dexmo. And we are super excited to share them with you since we finally got all patent applications filed...
Obviously hands are important. In virtual reality, seeing your hands gives you better immersion, and it allows you to truly interact with the virtual world: this is just the way it should be, not keyboards or mouse. But what's keeping us from embracing the digital world where we could see our hands? We believe it's the price and cost of data gloves.
According to our observation, most commercialized data-golves uses IMU(i.e. ControlVR, Noitom neurons) or bend-sensors(I.e.5DT, cyber gloves). Bend-sensors are expensive, and usually to regenerate a good hand model, the data-glove needs to have at least 9 bend-sensors(usually 14). The pure hardware cost of a bend sensor solution glove is over $200, and they are selling it for more than 5 times of that price. IMU is slightly cheaper, but IMU modules produce less noises in raw data generally cost higher. And that's probably what's keeping the price high.
Anyway, we figured out what's written above one year ago and have been thinking if their is a another way. If we do it differently, if we don't start from IMU or bend sensors, can we make it cheaper? And we found the answer: Mechanical Exoskeleton!!! I can't insert image here so here is the link to our prototypes:
Early Sketch:
http://i.imgur.com/eDxJj9D.jpg
Dexmo Original:
http://i.imgur.com/F0fvUJJ.jpg
Dexmo Alpha:
http://i.imgur.com/eMDlp2H.jpg
Dexmo Beta:
http://i.imgur.com/o9rIC5K.jpg
Please don't mind about the exposed wires, and how "not beautiful" they are(we get this comment a lot, but honestly, I love how they look).in the consumer version they will be fixed...
Please allow me to quickly explain 1.how it works and 2.why it can be made cheaper than all other methods and 3.why exoskeleton shows a better potential for force control feedback.
1.When the users move their fingers, the rod link structure would follow the user’s finger movement mechanically, then the movement would be reflected onto the rotational sensor of Dexmo and will be sent wirelessly to a computer via the on-board Bluetooth module. Our SDK will take the raw data, run them through our special kinematics hand regeneration algorithms, and then a skeletal hand model will be shown.
2.Noted we transform the bending of the finger into a rotational movement. And rotational sensors are stable, accessible and cheap.. I mean really cheap, a super high quality rotational sensor cost less than 2 dollars. All mechanical linkage can be made with injection molding, and in bulk quantity the plastic parts can be made really inexpensively.
3."There is more!", mechanical data gloves, due to its exoskeletonic nature, are very suitable for adding force-feedback systems. Our team is currently working on a lightweight, compact force-feedback systems that can be added easily on to dexmo. We will show you guys the pictures and videos when we made some huge progress.
We will put the detailed videos up later, now we only have one demo combo on the website showing some of our early demos. We are very excited about this technology, we spent a lot of time on it, and I hope you guys can like this as well!
Ok I need some sleep, it's 5 am in my current timezone...See you guys later, if you have any questions, feel free to ask, I will answer all of them when I wake up. If you can help spread the news for us, thousand of thanks, we would love to share this with more people!
Obviously hands are important. In virtual reality, seeing your hands gives you better immersion, and it allows you to truly interact with the virtual world: this is just the way it should be, not keyboards or mouse. But what's keeping us from embracing the digital world where we could see our hands? We believe it's the price and cost of data gloves.
According to our observation, most commercialized data-golves uses IMU(i.e. ControlVR, Noitom neurons) or bend-sensors(I.e.5DT, cyber gloves). Bend-sensors are expensive, and usually to regenerate a good hand model, the data-glove needs to have at least 9 bend-sensors(usually 14). The pure hardware cost of a bend sensor solution glove is over $200, and they are selling it for more than 5 times of that price. IMU is slightly cheaper, but IMU modules produce less noises in raw data generally cost higher. And that's probably what's keeping the price high.
Anyway, we figured out what's written above one year ago and have been thinking if their is a another way. If we do it differently, if we don't start from IMU or bend sensors, can we make it cheaper? And we found the answer: Mechanical Exoskeleton!!! I can't insert image here so here is the link to our prototypes:
Early Sketch:
http://i.imgur.com/eDxJj9D.jpg
Dexmo Original:
http://i.imgur.com/F0fvUJJ.jpg
Dexmo Alpha:
http://i.imgur.com/eMDlp2H.jpg
Dexmo Beta:
http://i.imgur.com/o9rIC5K.jpg
Please don't mind about the exposed wires, and how "not beautiful" they are(we get this comment a lot, but honestly, I love how they look).in the consumer version they will be fixed...
Please allow me to quickly explain 1.how it works and 2.why it can be made cheaper than all other methods and 3.why exoskeleton shows a better potential for force control feedback.
1.When the users move their fingers, the rod link structure would follow the user’s finger movement mechanically, then the movement would be reflected onto the rotational sensor of Dexmo and will be sent wirelessly to a computer via the on-board Bluetooth module. Our SDK will take the raw data, run them through our special kinematics hand regeneration algorithms, and then a skeletal hand model will be shown.
2.Noted we transform the bending of the finger into a rotational movement. And rotational sensors are stable, accessible and cheap.. I mean really cheap, a super high quality rotational sensor cost less than 2 dollars. All mechanical linkage can be made with injection molding, and in bulk quantity the plastic parts can be made really inexpensively.
3."There is more!", mechanical data gloves, due to its exoskeletonic nature, are very suitable for adding force-feedback systems. Our team is currently working on a lightweight, compact force-feedback systems that can be added easily on to dexmo. We will show you guys the pictures and videos when we made some huge progress.
We will put the detailed videos up later, now we only have one demo combo on the website showing some of our early demos. We are very excited about this technology, we spent a lot of time on it, and I hope you guys can like this as well!
Ok I need some sleep, it's 5 am in my current timezone...See you guys later, if you have any questions, feel free to ask, I will answer all of them when I wake up. If you can help spread the news for us, thousand of thanks, we would love to share this with more people!