Note: This text is a work in progress. If you find any inaccuracies or material that needs to be included, please contact us.
The LEGO hand shown here for the first time in April has undergone some improvement. Advancements of note include articulation and the wrist now moves naturally. I have not yet acquired any motors to actually move the device, but the hand will function by hand. I will be tearing it down at some point here to facilitate making instructions.
The articulation is done with zip ties. After extensive testing with cables and wire clamps, I thought I would mimic nature. The result has been very good as the zip ties act very much like real tendons.
The wrist functions on the same principle as a real one. The two support bones twist from an axis on the elbow joint. The next adventure will be figuring out to power it!
Check out the video below!
John Bergmann has been hard at work on the LEGO hand project. Check out his video update (now finally available embedded through Google Video):
One of the interests of the Open Prosthetics Project is to develop low cost prototyping platforms to encourage experimentation and collaboration on new concepts, both for control systems and mechanisms. One of our favorite prototyping tools is the LEGO construction system. We think that LEGO parts can offer a flexible and inexpensive method for developing complete mechatronic devices for research and development.
John Bergmann has made great progress in developing an advanced mechanism for a LEGO hand, complete with fully articulated fingers and opposable thumbs at a scale similar to that of the average male hand. He has put together some great documentation of his construction techniques for the basic hand.
At present this hand is not powered, however, John has been working a next revision of the device that includes a functional wrist with cable actuators. We encourage others to look at John’s design and consider collaborating with him to continue the development of this platform. Once a functional power system is integrated the device could serve as an excellent plant to be controlled by a low-cost myoelectric or acoustic myography sensor system.
Gluing a structure like this together could make it strong enough even for limited use. Experts say that Oatey All-Purpose Glue works well. We have yet to give it a try, although we can say that it can be frustrating to have a model come apart when you try to use it.
Development of a LEGO-based sensor platform is another area of interest. We believe that a complete, low-cost myoelectric sensor system could be built using LEGO Minstorms NXT controllers. Creating an NXT-based sensor platform combined with a hand mechanism such as John’s would result in a complete LEGO-based reference design for less than $500 in materials.
Given the small number of upper-extremity prosthetic clients, a toy that uses myoelectric technology could spur some interesting developments. Also, LEGO products are accessible in ways that many things in the world of hardware are not. The existence of a complete myoelectric platform like this would significantly lower barriers to participation in research and development and aid in accelerating collaboration by using inexpensive, readily available parts for implementation.
Here you can download the 3D parts for assembling the LEGO hand.
All the parts are in PROE Wildfire 4.0 format in order, if you need an specific format please don’t hesitate to ask me and I’ll gladly send the needed files that will be compatible with the CAD software that you may be using.
Esteban Alvarez Serrano
Mechanical Design Engineer
52 (664) 183 1190
The only type of mechatronic prosthesis commercially available today is the myoelectric prosthesis. Myoelectric protheses are controlled by tiny voltages generated on the surface of the skin by the activity of residual muscles (Surface Electromyelogram (EMG)). These voltages are amplified and processed so that when the user flexes a muscle, motors on the prosthesis move in a predictable way. A myoelectric prosthesis does away with the harness and cables of a body-powered device and can be made to look very natural, although at the cost of a more restrictive and less comfortable socket. Another downside to these devices is their lack of responsiveness; most hands move slowly to increase grip force or battery life at the expense of fast action. However, the major barrier to adoption for most people is money: an arm costs at least $30,000.
We are in the process of researching surface EMG to discover the factors that drive the price of the electronics. If you have expertise in the field, please contact us.
Along with EMG, we are considering the use of several other technologies which may offer reduced cost and increased reliability.
Electrical Impedance Tomography (EIT)
Impedance tomography is a method of getting a cross-sectional image of the body very safely and cheaply. All that is needed is a ring of electrodes, often as few as 16, contacting the skin, a source of constant current, and an analog to digital converter to sense voltages on the electrodes. A small alternating current is passed through two of the electrodes and the voltage is recorded at all of the others. This process is then repeated very rapidly while changing the electrode pair the current passes through. The resulting data is then used to solve a non-linear system of equations, producing an image of the tissues between the electrodes. Different tissues resist electric current to different degrees, and so muscle, bone, and blood can all be differentiated with this method. Further details of the process can be found in chapter 26 of the web edition of Bioelectromagnetism. You can also look at some sample images and videos of EIT data. Possible analyses of the data might include changes in the cross-sectional area of a contracted muscle or changes in the impedance of a contracted muscle.
Advantages of EIT:
- It generates electricity for measurement, and so is not as susceptible to noise as EMG.
- The redundancy of multiple sensors might make it less sensitive to noise and contact positioning than EMG.
- It may require substantially less analog amplification and filtering than EMG.
Disadvantages of EIT:
- It probably requires an array of many sensors, which in turn will need multiplexing and lots of labor-intensive freeform wiring.
- It will need some computing power to handle the data stream, but this may be less than expected because a full and accurate image is not needed by a prosthetic control system.
- It will require a constant supply of power, at least a few milliamps, which will use battery life.
- To our knowledge, this techology is untested in prosthetics and is still mostly in the academic realm.
Acoustic/Mechanical Myography (AMG/MMG)
Muscles actually make sounds when contracting, which you can test by putting your thumbs in your ears and making a fist. The low rumbling sound you hear is your muscle fibers resonating as they contract. It turns out that this resonance (at between 20 and 30 Hz) can be picked up by a subsonic microphone and easily analyzed to detect muscle movements. The power (loudness) of the sound correlates well with the force on the muscle. The phenomenon is described in this article, and its use in prosthetics is covered by US patent 4571750, which will expire in February 2006, but also by US patent 4748987, which will not expire until June of 2008. We are attempting to get in touch with the inventor, Daniel T. Barry to discuss intellectual property issues.
Advantages of MMG:
- Muscle sounds can be monitored by cheap microphones, even through a prosthetic sock, without many of the concerns of fitting an electrode.
- The signal is more powerful than EMG, and less susceptible to electrical interference, so only sound shielding is needed.
- The signals can be processed by ubiquitous and cheap digital audio hardware like DSPs.
Disadvantages of MMG:
- It detects actual muscle effort, not intent, so fatigue will tend to reduce the response of the prosthesis.
- It is susceptible to loud environmental noises. This effect can probably be mitigated by active noise control, the same technology used in some consumer headphones.
It looks like mechanomyography (MMG) is the right technology to use, since we have found a research group that has demonstrated its practicality. Jorge Silva, Winfried Haim, and Tom Chau of the PRISM Lab have made and tested a self-contained MMG controlled prosthesis1. Their device used three sensors, each consisting of a microphone to pick up muscle sounds and an accelerometer to detect external interference. The sensors were integrated into a soft silicone socket at the end of the user’s residual limb and placed 120 degrees apart. Their signals were interpreted by a microcontroller and converted into simulated EMG signals to control an Otto Bock hand. Their results were promising, and the devices performed with over 70% control accuracy. The prosthesis was difficult to control when the user was sustaining heavy loads with it, when the arm was held over the head, and when the user was walking. However, none of the usual problems with EMG electrode placement showed up. Active Living Magazine has published an article on the project.
It seems like major improvements can be made on this design if it were built from the ground up to be controlled by MMG signals. Better use might be made of the sensor data, the possible unreliability of the EMG control electronics could be eliminated, and the latency might be reduced from the paper’s average of 120ms.
1 Silva, J., Heim, W., & Chau, T. (2005).“A self-contained mechanomyography-driven externally powered prosthesis”, Archives of Physical Medicine and Rehabilitation, 86(10):2066-2070.
We have built a very simple data collection system for recording MMG signals (muscle sounds). It consists of a 3.5mm audio plug, a Panasonic microphone cartridge (part number WM-64PNT), some wire, and a laptop. Not including the laptop, the components cost less than $5. The microphone is an electret condenser type, and is powered by the laptop’s sound card. We used the free program Audacity to record and process the sound. We simply taped the microphone directly onto the skin with masking tape, plugged it into the laptop, and pressed record:
The results were much better than we expected, especially after we rolled Jon’s silicone suspension sock over the microphone for isolation, which made the sound much louder. We processed the data to make it more audible by filtering out all frequencies above 40Hz and raising the pitch by two octaves. Here’s an mp3 of the processed sound. In this recording, Jon is making the motions that would have opened and closed his hand. The microphone is over the muscle that opens his hand, so you can hear it loudest, but the muscle contracting on the other side is also audible. You can hear the muscle sounds getting louder over the course of the recording as he grips with progressively more force. You can download the raw data as a WAV file (1.5M) if you want to do your own processing (the popping noises are from the alligator clips connecting the microphone to the computer).
After some more tests with electret condenser microphones, we have found out some essential characteristics of an ECM based MMG sensor:
- Clean power supply – Just running our laptop on batteries made a huge difference in low-frequency noise. Obviously a prosthetic device would be battery powered, but we need to protect against power supply fluctuations from the motors, etc.
- No electronics-body contact – When the metal case of the microphone touches the body, it turns into a giant antenna, and the 60Hz radiation from household electrical power really interferes with our signal.
- Good sealing – There must be an airtight chamber between the skin and the microphone or a the signal gets dramatically quieter. Additionally, the size and shape of the air cavity plays a big role in the sensitivity of the sensor.
The first requirement is easily satisfied with electronic filtering and the second simply requires insulation and shielding. However, the requirement of a sealed air chamber imposes some significant manufacturing costs. Jorge Silva’s CMASP sensor units require two types of silicone cast around the microphone. Silicone casting in a small shop is very labor intensive and time consuming. The sealing requirement would be completely eliminated if we did not use air to conduct sound from the skin to the sensor. We are investigating alternative sensor types and have discovered the following:
- Piezoelectric film elements – These are thin sheets of material that produce a voltage when bent. An element costs from $5 to $15 in low volume depending on options from Measurement Specialties, Inc.. The chief advantage of these sensors is that they are low-profile and don’t contain easily broken connections. It seems like they could easily be integrated into a silicone suspension sleeve, especially the type that has integrated, flexible leads. Here is a technical discussion of piezoelectric films in general. An excerpt: “One major advantage of piezo film over piezo ceramic is its low acoustic impedance which is closer to that of water, human tissue and other organic materials . . . [which] permits more efficient transduction of acoustic signals in water and tissue.”
- Accelerometers – These are accelerometers that are mounted directly on a circuit board. From MSI they cost $20 for a one-axis chip. The disadvantage of this solution is that the measurement is taken from movement of the whole device, whether it is moved by muscle vibrations or by the whole arm being shaken. It also requires surface mount soldering and a circuit board built into the suspension sleeve.
- Vibration transducers and contact microphones – These are encapsulated devices that measure vibrations when mounted to a surface. One from Knowles Acoustics costs around $50 from DigiKey.
The piezoelectric film elements seem really promising, so we are ordering some to test out. The next step is to collect data from more than one microphone at once and begin testing some analysis methods.
All content and designs on this site are in the public domain, and we place no restrictions on their use. We encourage any derivative works, but all designs are registered periodically so that our work cannot be kept from the public by patents.