Robot Learns To Smile And Frown

SOURCE

ScienceDaily (July 8, 2009) — A hyper-realistic Einstein robot at the University of California, San Diego has learned to smile and make facial expressions through a process of self-guided learning. The UC San Diego researchers used machine learning to “empower” their robot to learn to make realistic facial expressions.

“As far as we know, no other research group has used machine learning to teach a robot to make realistic facial expressions,” said Tingfan Wu, the computer science Ph.D. student from the UC San Diego Jacobs School of Engineering who presented this advance on June 6 at the IEEE International Conference on Development and Learning.
The faces of robots are increasingly realistic and the number of artificial muscles that controls them is rising. In light of this trend, UC San Diego researchers from the Machine Perception Laboratory are studying the face and head of their robotic Einstein in order to find ways to automate the process of teaching robots to make lifelike facial expressions.
This Einstein robot head has about 30 facial muscles, each moved by a tiny servo motor connected to the muscle by a string. Today, a highly trained person must manually set up these kinds of realistic robots so that the servos pull in the right combinations to make specific face expressions. In order to begin to automate this process, the UCSD researchers looked to both developmental psychology and machine learning.
Developmental psychologists speculate that infants learn to control their bodies through systematic exploratory movements, including babbling to learn to speak. Initially, these movements appear to be executed in a random manner as infants learn to control their bodies and reach for objects.
“We applied this same idea to the problem of a robot learning to make realistic facial expressions,” said Javier Movellan, the senior author on the paper presented at ICDL 2009 and the director of UCSD’s Machine Perception Laboratory, housed in Calit2, the California Institute for Telecommunications and Information Technology.
Although their preliminary results are promising, the researchers note that some of the learned facial expressions are still awkward. One potential explanation is that their model may be too simple to describe the coupled interactions between facial muscles and skin.
To begin the learning process, the UC San Diego researchers directed the Einstein robot head (Hanson Robotics’ Einstein Head) to twist and turn its face in all directions, a process called “body babbling.” During this period the robot could see itself on a mirror and analyze its own expression using facial expression detection software created at UC San Diego called CERT (Computer Expression Recognition Toolbox). This provided the data necessary for machine learning algorithms to learn a mapping between facial expressions and the movements of the muscle motors.
Once the robot learned the relationship between facial expressions and the muscle movements required to make them, the robot learned to make facial expressions it had never encountered.
For example, the robot learned eyebrow narrowing, which requires the inner eyebrows to move together and the upper eyelids to close a bit to narrow the eye aperture.
“During the experiment, one of the servos burned out due to misconfiguration. We therefore ran the experiment without that servo. We discovered that the model learned to automatically compensate for the missing servo by activating a combination of nearby servos,” the authors wrote in the paper presented at the 2009 IEEE International Conference on Development and Learning.
“Currently, we are working on a more accurate facial expression generation model as well as systematic way to explore the model space efficiently,” said Wu, the computer science PhD student. Wu also noted that the “body babbling” approach he and his colleagues described in their paper may not be the most efficient way to explore the model of the face.
While the primary goal of this work was to solve the engineering problem of how to approximate the appearance of human facial muscle movements with motors, the researchers say this kind of work could also lead to insights into how humans learn and develop facial expressions.
“Learning to Make Facial Expressions,” by Tingfan Wu, Nicholas J. Butko, Paul Ruvulo, Marian S. Bartlett, Javier R. Movellan from Machine Perception Laboratory, University of California San Diego. Presented on June 6 at the 2009 IEEE 8th International Conference On Development And Learning.
Adapted from materials provided by University of California - San Diego.

Robo-bats With Metal Muscles May Be Next Generation Of Remote Control Flyers

SOURCE

ScienceDaily (July 7, 2009) — Tiny flying machines can be used for everything from indoor surveillance to exploring collapsed buildings, but simply making smaller versions of planes and helicopters doesn't work very well. Instead, researchers at North Carolina State University are mimicking nature's small flyers – and developing robotic bats that offer increased maneuverability and performance.

Small flyers, or micro-aerial vehicles (MAVs), have garnered a great deal of interest due to their potential applications where maneuverability in tight spaces is necessary, says researcher Gheorghe Bunget. For example, Bunget says, "due to the availability of small sensors, MAVs can be used for detection missions of biological, chemical and nuclear agents." But, due to their size, devices using a traditional fixed-wing or rotary-wing design have low maneuverability and aerodynamic efficiency.
So Bunget, a doctoral student in mechanical engineering at NC State, and his advisor Dr. Stefan Seelecke looked to nature. "We are trying to mimic nature as closely as possible," Seelecke says, "because it is very efficient. And, at the MAV scale, nature tells us that flapping flight – like that of the bat – is the most effective."
The researchers did extensive analysis of bats' skeletal and muscular systems before developing a "robo-bat" skeleton using rapid prototyping technologies. The fully assembled skeleton rests easily in the palm of your hand and, at less than 6 grams, feels as light as a feather. The researchers are currently completing fabrication and assembly of the joints, muscular system and wing membrane for the robo-bat, which should allow it to fly with the same efficient flapping motion used by real bats.
"The key concept here is the use of smart materials," Seelecke says. "We are using a shape-memory metal alloy that is super-elastic for the joints. The material provides a full range of motion, but will always return to its original position – a function performed by many tiny bones, cartilage and tendons in real bats."
Seelecke explains that the research team is also using smart materials for the muscular system. "We're using an alloy that responds to the heat from an electric current. That heat actuates micro-scale wires the size of a human hair, making them contract like 'metal muscles.' During the contraction, the powerful muscle wires also change their electric resistance, which can be easily measured, thus providing simultaneous action and sensory input. This dual functionality will help cut down on the robo-bat's weight, and allow the robot to respond quickly to changing conditions – such as a gust of wind – as perfectly as a real bat."
In addition to creating a surveillance tool with very real practical applications, Seelecke says the robo-bat could also help expand our understanding of aerodynamics. "It will allow us to do tests where we can control all of the variables – and finally give us the opportunity to fully understand the aerodynamics of flapping flight," Seelecke says.
Bunget will present the research this September at the American Society of Mechanical Engineers Conference on Smart Materials, Adaptive Structures and Intelligent Systems in Oxnard, Calif.
Adapted from materials provided by North Carolina State University.

Quadriplegics Can Operate Powered Wheelchair With Tongue Drive System

SOURCE

ScienceDaily (July 6, 2009) — An assistive technology that enables individuals to maneuver a powered wheelchair or control a mouse cursor using simple tongue movements can be operated by individuals with high-level spinal cord injuries, according to the results of a recently completed clinical trial.

"This clinical trial has validated that the Tongue Drive system is intuitive and quite simple for individuals with high-level spinal cord injuries to use," said Maysam Ghovanloo, an assistant professor in the School of Electrical and Computer Engineering at the Georgia Institute of Technology. "Trial participants were able to easily remember and correctly issue tongue commands to play computer games and drive a powered wheelchair around an obstacle course with very little prior training."
At the annual conference of the Rehabilitation Engineering and Assistive Technology Society of North America (RESNA) on June 26, the researchers reported the results of the first five clinical trial subjects to use the Tongue Drive system. The trial was conducted at the Shepherd Center, an Atlanta-based catastrophic care hospital, and funded by the National Science Foundation and the Christopher and Dana Reeve Foundation.
The clinical trial tested the ability of these individuals with tetraplegia, as a result of high-level spinal cord injuries (cervical vertebrae C3-C5), to perform tasks related to computer access and wheelchair navigation -- using only their tongue movements.
At the beginning of each trial, Ghovanloo and graduate students Xueliang Huo and Chih-wen Cheng attached a small magnet -- the size of a grain of rice -- to the participant's tongue with tissue adhesive. Movement of this magnetic tracer was detected by an array of magnetic field sensors mounted on wireless headphones worn by the subject. The sensor output signals were wirelessly transmitted to a portable computer, which was carried on the wheelchair.
The signals were processed to determine the relative motion of the magnet with respect to the array of sensors in real-time. This information was then used to control the movements of the cursor on a computer screen or to substitute for the joystick function in a powered wheelchair. Details on use of the Tongue Drive for wheeled mobility were published in the June 2009 issue of the journal IEEE Transactions on Biomedical Engineering.
Ghovanloo chose the tongue to operate the system because unlike hands and feet, which are controlled by the brain through the spinal cord, the tongue is directly connected to the brain by a cranial nerve that generally escapes damage in severe spinal cord injuries or neuromuscular diseases.
Before using the Tongue Drive system, the subjects trained the computer to understand how they would like to move their tongues to indicate different commands. A unique set of specific tongue movements was tailored for each individual based on the user's abilities, oral anatomy and personal preferences. For the first computer test, the user issued commands to move the computer mouse left and right. Using these commands, each subject played a computer game that required moving a paddle horizontally to prevent a ball from hitting the bottom of the screen.
After adding two more commands to their repertoire -- up and down -- the subjects were asked to move the mouse cursor through an on-screen maze as quickly and accurately as possible.
Then the researchers added two more commands -- single and double mouse clicks -- to provide the subject with complete mouse functionality. When a randomly selected symbol representing one of the six commands appeared on the computer screen, the subject was instructed to issue that command within a specified time period. Each subject completed 40 trials for each time period.
After the computer sessions, the subjects were ready for the wheelchair driving exercise. Using forward, backward, right, left and stop/neutral tongue commands, the subjects maneuvered a powered wheelchair through an obstacle course.
The obstacle course contained 10 turns and was longer than a professional basketball court. Throughout the course, the users had to perform navigation tasks such as making a U-turn, backing up and fine-tuning the direction of the wheelchair in a limited space. Subjects were asked to navigate through the course as fast as they could, while avoiding collisions.
Each subject operated the powered wheelchair using two different control strategies: discrete mode, which was designed for novice users, and continuous mode for more experienced users. In discrete mode, if the user issued the command to move forward and then wanted to turn right, the user would have to stop the wheelchair before issuing the command to turn right. The stop command was selected automatically when the tongue returned to its resting position, bringing the wheelchair to a standstill.
"Discrete mode is a safety feature particularly for novice users, but it reduces the agility of the wheelchair movement," explained Ghovanloo. "In continuous mode, however, the user is allowed to steer the powered wheelchair to the left or right as it is moving forward and backward, thus making it possible to follow a curve."
Each subject completed the course at least twice using each strategy while the researchers recorded the navigation time and number of collisions. Using discrete control, the average speed for the five subjects was 5.2 meters per minute and the average number of collisions was 1.8. Using continuous control, the average speed was 7.7 meters per minute and the average number of collisions was 2.5.
While this initial performance trial only required six tongue commands, the Tongue Drive system can potentially capture a large number of tongue movements, each of which can represent a different user command. The ability to train the system with as many commands as an individual can comfortably remember and having all of the commands available to the user at the same time are significant advantages over the common sip-n-puff device that acts as a simple switch controlled by sucking or blowing through a straw.
Some sip-n-puff users also consider the straw to be a symbol of their disability. Since Tongue Drive users simply wear headphones that are commonly worn to listen to music, the system is more acceptable to potential users.
John Anschutz, manager of the assistive technology program at the Shepherd Center, identified advantages the Tongue Drive system has over the tongue-touch keypad.
"The Tongue Drive system seems to be much more supportable if there were a failure of some component within the system. With the old tongue-touch keypad, if the system went down then the user lost all of the functions of the wheelchair, phone, computer and environmental control," explained Anschutz. "Ghovanloo's approach should be much more repairable should a fault arise, which is critical for systems for which so much function is depended upon."
A future system upgrade will be to move the sensors inside the user's mouth, according to Ghovanloo. This will be an important step for users who are very impaired and cannot reposition the system for best results, according to Anschutz.
"All of the subjects successfully completed the computer and powered wheelchair navigation tasks with their tongues without difficulty, which demonstrates that the Tongue Drive system can potentially provide individuals unable to move their arms and hands with effective control over a wide variety of devices they use in their daily lives," said Ghovanloo.
Adapted from materials provided by Georgia Institute of Technology, via EurekAlert!, a service of AAAS.

Robot Soccer: Cooperative Soccer Playing Robots Compete

SOURCE

ScienceDaily (July 6, 2009) — The cooperative soccer playing robots of the Universität Stuttgart are world champions in the middle size league of robot soccer. After one of the most interesting competitions in the history of Robocup from 29th June to 5th July, 2009, in Graz, the 1. RFC Stuttgart on the last day of the competition succeeded in winning the world championship 2009 in an exciting game against the team of Tech United from Eindhoven (The Netherlands) with the final result of 4:1.

During the competition Stuttgart's robots had to make their way against 13 other teams from eight countries, among them the current world champion Cambada (Portugal). Besides the teams from Germany, Italy, The Netherlands, Portugal, and Austria, teams from China, Japan, and Iran competed against each other.
The 1.RFC Stuttgart includes staff of two Institutes, namely the department of Image Understanding (Head: Prof. Levi) of the Institute of Parallel and Distributed Systems and the Institute of Technical Optics (Head: Prof. Osten), achieved also the 2nd place at the so-called "technical challenge" and a further 1st place at the "scientific challenge".
After the final match of the competition, the middle-size league robots of the 1. RFC Stuttgart - the new world champion - had to play against the human officials of the RoboCup federation. It turned out, that hereby the robots were the inferior team. Clearly the RoboCup community has still to bridge a vast distance to reach their final goal to let a humanoid robot team play against the human world champion by the year 2050.
The success tells its own tale but one might wonder which scientific interest is behind the RoboCup competitions. Preconditions for the successful participation at these competitions are extensive efforts in current research topics of computer science such as real-time image processing and architectures, cooperative robotics and distributed planning. Possible application scenarios of these research activities reach from autonomous vehicles, cooperative manufacturing robotics, service robotics to the point of planetary or deep-sea exploration by autonomous robotic systems. In this context autonomous means that no or only a limited human intervention is necessary.
Adapted from materials provided by University of Stuttgart.

Ultrasensitive Detector Promises Improved Treatment Of Viral Respiratory Infections

SOURCE

ScienceDaily (July 6, 2009) — A Vanderbilt chemist and a biomedical engineer have teamed up to develop a respiratory virus detector that is sensitive enough to detect an infection at an early stage, takes only a few minutes to return a result and is simple enough to be performed in a pediatrician's office.

Writing in The Analyst – a journal published by the Royal Society of Chemistry – the developers report that their technique, which uses DNA hairpins attached to gold filaments, can detect the presence of respiratory syncytial virus (RSV) – a leading cause of respiratory infections in infants and young children – at substantially lower levels than the standard laboratory assay.
"We hope that our research will help us break out of the catch-22 that is holding back major advances in the treatment of respiratory viruses," says Associate Professor of Chemistry David Wright, who is working with Professor of Biomedical Engineering Frederick "Rick" Haselton on the new detection method.
According to the chemist, major pharmaceutical companies are not investing in the development of antiviral drugs for RSV and the other major respiratory viruses because there is no way to detect the infections early enough for the drugs to work effectively without harmful side-effects. "There are antiviral compounds out there – we have discovered some of them in my lab – that would work if we can detect the virus early enough, before there is too much virus in the system," he says.
In addition, the lack of a reliable early detection system adds to the growing problem of antibiotic resistance. The symptoms of respiratory infections caused by viral agents are nearly identical to those caused by bacteria. As a result, antibiotics, which target bacteria, are often incorrectly prescribed for viral infections. Not only is this ineffective, but it also increases the number of antibiotic-resistant strains.
Currently, there are several standard tests for RSV including culturing the virus, polymerase chain reaction (PCR) and the enzyme-linked immunosorbent assay (ELISA). To have any of these tests done, doctors must send a mucous sample from a patient to a special laboratory. When combined with delivery times, backlogs and other delays, it frequently takes a day or more to get the results. Unfortunately, respiratory viruses multiply so rapidly that this can be too late for antiviral drugs to work, Wright says.
By contrast, "our system could easily be packaged in a disposable device about the size of a ballpoint pen," says Haselton. To perform a test, all that would be required is to pull off a cap that will expose a length of gold wire, dip the wire in the sample, pull the wire through the device and put the exposed wire into a fluorescence scanner. If it lights up, then the virus is present.
The new detector design is a combination of two existing technologies.
One is the filament-based antibody recognition assay (FARA) developed several years ago by Haselton and patented by Vanderbilt. FARA uses antibodies – special proteins produced by the immune system that binds to specific foreign substances – that are coated on the surface of a polyester filament. When the coated filament is exposed to a sample, if it contains any of the target molecules, they stick to the antibodies, forming complexes that can be detected with fluorescent dyes. One advantage of this approach is that a sample can be put through different processing steps simply by pulling the filament through a series of small chambers. In the RSV detection application, the chambers contain washing solutions that remove non-specific binding molecules.
"Originally we thought that we would have to put special seals between the chambers but we found that if we make the openings small enough, then the solutions in the chambers stay in place as we pull the wire through," says Haselton.
The second technology is based on molecular beacon probes, an approach often used in PCR. The probes consist of short lengths of single-strand DNA that normally form a hairpin shape but straighten out when they are bound to a target molecule. A fluorescent dye molecule is attached to one leg of the hairpin and a molecule that quenches its fluorescence is attached to the other. When the probe is in its hairpin configuration, the dye and quencher molecules lay side by side so the probe does not fluoresce. When it is bound to a target, such as a piece of viral RNA, the ends spring apart, turning on the probe's fluorescence.
The Vanderbilt researchers realized that if they attached molecular beacons to a gold-coated filament, the gold could theoretically replace the quencher molecule and inhibit the beacon's fluorescence. However, they had to find a linking molecule – the molecule that attaches the beacon to the wire – that was just the right length to make it work.
Once they solved this problem, the researchers tested the sensitivity of the new system. They found that it could detect the presence of RSV virus particles at levels that are 200 times below the minimum detection level of the standard ELISA method. This extreme sensitivity combined with the basic simplicity of the approach makes it "attractive for further development as a viral detection platform," the scientists write in the Analyst article, which was published online May 15.
According to Haselton, there are two areas where further development is needed. One is sample preparation. Commercial RNA sample preparation kits are available, but they are more expensive and complex than desirable. The team is currently examining the design of a simple pull-through RNA isolation chamber. The team is also exploring ways to reduce false detections. There are a lot of other molecules in mucous besides viral RNA that can bind to some extent with the molecular beacons. However, the researchers argue that it should be possible to reduce the number of false positives significantly by adding a heating step that is calibrated to drive off the molecules that are less strongly bound to the beacons than the viral RNA.
The next major step in the development process is to see how the device performs with real patient samples.
This research was supported by grants from Vanderbilt University and the National Institutes of Health.
Adapted from materials provided by Vanderbilt University.

Innovative Technology Shatters The Barriers Of Modern Light Microscopy

SOURCE

ScienceDaily (July 5, 2009) — Researchers at the Helmholtz Zentrum München and the Technische Universität München are using a combination of light and ultrasound to visualize fluorescent proteins that are seated several centimeters deep into living tissue. In the past, even modern technologies have failed to produce high-resolution fluorescence images from this depth because of the strong scattering of light.

In the Nature Photonics journal, the Munich researchers describe how they can reveal genetic expression within live fly larvae and fish by “listening to light”. In the future this technology may facilitate the examination of tumors or coronary vessels in humans.
Since the dawn of the microscope scientists have been using light to scrutinize thin sections of tissue to ascertain whether they are healthy or diseased or to investigate cell function. However, the penetration limits for this kind of examination lie between half a millimeter and one millimeter of tissue. In thicker layers light is diffused so strongly that all useful details are obscured.
Together with his research team, Professor Vasilis Ntziachristos, director of the Institute of Biological and Medical Imaging of the Helmholtz Zentrum München – German Research Center for Environmental Health and chair for biological imaging at the Technische Universität München, has now broken through this barrier and rendered three-dimensional images through at least six millimeters of tissue, allowing whole-body visualization of adult zebra fish.
To achieve this feat, Prof. Ntziachristos and his team made light audible. They illuminated the fish from multiple angles using flashes of laser light that are absorbed by fluorescent pigments in the tissue of the genetically modified fish. The fluorescent pigments absorb the light, a process that causes slight local increases temperature, which in turn result in tiny local volume expansions. This happens very quickly and creates small shock waves. In effect, the short laser pulse gives rise to an ultrasound wave that the researchers pick up with an ultrasound microphone.
The real power of the technique, however, lies in specially developed mathematical formulas used to analyze the resulting acoustic patterns. An attached computer uses these formulas to evaluate and interpret the specific distortions caused by scales, muscles, bones and internal organs to generate a three-dimensional image.
The result of this “multi-spectral opto-acoustic tomography”, or MSOT, is an image with a striking spatial resolution better than 40 micrometers (four hundredths of a millimeter). And best of all, the sedated fish wakes up and recovers without harm following the procedure.
Dr. Daniel Razansky, who played a pivotal role in developing the method, says, "This opens the door to a whole new universe of research. For the first time, biologists will be able to optically follow the development of organs, cellular function and genetic expression through several millimeters to centimeters of tissue.”
In the past, understanding the evolution of development or of disease required numerous animals to be sacrificed. With a plethora of fluorochrome pigments to choose from – including pigments using the fluorescence protein technology for which a Nobel Prize was awarded in 2008 and clinically approved fluorescent agents – observing metabolic and molecular processes in all kinds of living organisms, from fish to mice and humans, will be possible. The fruits of pharmaceutical research can also be harvested faster since the molecular effects of new treatments can be observed in the same animals over an extended period of time.
Bio-engineer Ntziachristos is convinced that, “MSOT can truly revolutionize biomedical research, drug discovery and healthcare. Since MSOT allows optical and fluorescence imaging of tissue to a depth of several centimeters, it could become the method of choice for imaging cellular and subcellular processes throughout entire living tissues.”
Journal reference:
Razansky et al. Multispectral opto-acoustic tomography of deep-seated fluorescent proteins in vivo. Nature Photonics, 2009; 3 (7): 412 DOI: 10.1038/nphoton.2009.98
Adapted from materials provided by Helmholtz Zentrum München - German Research Center for Environmental Health.

Nanotechnology May Increase Longevity Of Dental Fillings

SOURCE

ScienceDaily (July 3, 2009) — Tooth-colored fillings may be more attractive than silver ones, but the bonds between the white filling and the tooth quickly age and degrade. A Medical College of Georgia researcher hopes a new nanotechnology technique will extend the fillings' longevity.

"Dentin adhesives bond well initially, but then the hybrid layer between the adhesive and the dentin begins to break down in as little as one year," says Dr. Franklin Tay, associate professor of endodontics in the MCG School of Dentistry. "When that happens, the restoration will eventually fail and come off the tooth."
Half of all tooth-colored restorations, which are made of composite resin, fail within 10 years, and about 60 percent of all operative dentistry involves replacing them, according to research in the Journal of the American Dental Association.
"Our adhesives are not as good as we thought they were, and that causes problems for the bonds," Dr. Tay says.
To make a bond, a dentist etches away some of the dentin's minerals with phosphoric acid to expose a network of collagen, known as the hybrid layer. Acid-etching is like priming a wall before it's painted; it prepares the tooth for application of an adhesive to the hybrid layer so that the resin can latch on to the collagen network. Unfortunately, the imperfect adhesives leave spaces inside the collagen that are not properly infiltrated with resin, leading to the bonds' failure.
Dr. Tay is trying to prevent the aging and degradation of resin-dentin bonding by feeding minerals back into the collagen network. With a two year, $252,497 grant from the National Institute of Dental & Craniofacial Research, he will investigate guided tissue remineralization, a new nanotechnology process of growing extremely small, mineral-rich crystals and guiding them into the demineralized gaps between collagen fibers.
His idea came from examining how crystals form in nature. "Eggshells and abalone [sea snail] shells are very strong and intriguing," Dr. Tay says. "We're trying to mimic nature, and we're learning a lot from observing how small animals make their shells."
The crystals, called hydroxyapatite, bond when proteins and minerals interact. Dr. Tay will use calcium phosphate, a mineral that's the primary component of dentin, enamel and bone, and two protein analogs also found in dentin so he can mimic nature while controlling the size of each crystal.
Crystal size is the real challenge, Dr. Tay says. Most crystals are grown from one small crystal into a larger, homogeneous one that is far too big to penetrate the spaces within the collagen network. Instead, Dr. Tay will fit the crystal into the space it needs to fill. "When crystals are formed, they don't have a definite shape, so they are easily guided into the nooks and crannies of the collagen matrix," he says.
In theory, the crystals should lock the minerals into the hybrid layer and prevent it from degrading. If Dr. Tay's concept of guided tissue remineralization works, he will create a delivery system to apply the crystals to the hybrid layer after the acid-etching process.
"Instead of dentists replacing the teeth with failed bonds, we're hoping that using these crystals during the bond-making process will provide the strength to save the bonds," Dr. Tay says. "Our end goal is that this material will repair a cavity on its own so that dentists don't have to fill the tooth."
Adapted from materials provided by Medical College of Georgia.