Elias S. Manolakos

Learning human reach-to-grasp strategies: Towards EMG-based control of robotic arm-hand systems

Reaching and grasping of objects in an everyday-life environment seems so simple for humans, though so complicated from an engineering point of view. Humans use a variety of strategies for reaching and grasping anything from the simplest to the most complicated objects, achieving high dexterity and efficiency. This seemingly simple process of reach-to-grasp relies on the complex coordination of the musculoskeletal system of the upper limbs. In this paper, we study the muscular co-activation patterns during a variety of reach-to-grasp motions, and we introduce a learning scheme that can discriminate between different strategies. This scheme can then classify reach-to-grasp strategies based on the muscular co-activations. We consider the arm and hand as a whole system, therefore we use surface ElectroMyoGraphic (sEMG) recordings from muscles of both the upper arm and the forearm. The proposed scheme is tested in extensive paradigms proving its efficiency, while it can be used as a switching mechanism for task-specific motion and force estimation models, improving EMG-based control of robotic arm-hand systems.

Improving 2-DE gel image denoising using contourlets

One of the most commonly used methods for protein separation is 2‐DE. After 2‐DE gel scanning, images with a plethora of spot features emerge that are usually contaminated by inherent noise. The objective of the denoising process is to remove noise to the extent that the true spots are recovered correctly and accurately i.e. without introducing distortions leading to the detection of false‐spot features. In this paper we propose and justify the use of the contourlet transform as a tool for 2‐DE gel images denoising. We compare its effectiveness with state‐of‐the‐art methods such as wavelets‐based multiresolution image analysis and spatial filtering. We show that contourlets not only achieve better average S/N performance than wavelets and spatial filters, but also preserve better spot boundaries and faint spots and alter less the intensities of informative spot features, leading to more accurate spot volume estimation and more reliable spot detection, operations that are essential to differential expression proteomics for biomarkers discovery.

A Learning Scheme for Reach to Grasp Movements: On EMG-Based Interfaces Using Task Specific Motion Decoding Models

A learning scheme based on random forests is used to discriminate between different reach to grasp movements in 3-D space, based on the myoelectric activity of human muscles of the upper-arm and the forearm. Task specificity for motion decoding is introduced in two different levels: Subspace to move toward and object to be grasped. The discrimination between the different reach to grasp strategies is accomplished with machine learning techniques for classification. The classification decision is then used in order to trigger an EMG-based task-specific motion decoding model. Task specific models manage to outperform “general” models providing better estimation accuracy. Thus, the proposed scheme takes advantage of a framework incorporating both a classifier and a regressor that cooperate advantageously in order to split the task space. The proposed learning scheme can be easily used to a series of EMG-based interfaces that must operate in real time, providing data-driven capabilities for multiclass problems, that occur in everyday life complex environments.

A P2P SOA enabling group collaboration through service composition

The Service-Oriented Architecture (SOA) paradigm was introduced for exposing business processes as services and enabling their interaction and composition over the Internet. The same computing model can potentially be extended to services of personal devices. As personal devices become network-aware their services can be made available (by their owners) to members of trusted peer groups, thus allowing them to compose new distributed collaborative applications. However, dealing with firewall traversals, NATs, mobility and issues of service-level naming and addressing stand in the way of this vision. In this paper we introduce a P2P SOA middleware architecture that addresses such problems of pervasive connectivity without requiring any intermediaries or changes to the service implementations. We present the basic elements of the architecture and the design of a specific instance of it, which enables the P2P service discovery and execution of composite personal services in distributed OSGi containers over JXTA.

IP-cores design for the kNN classifier

We present the systematic design of two linear array IP cores for the k-nearest neighbor (k-NN) benchmark classifier. The need for real-time classification of data vectors with possibly thousands of features (dimensions) motivates the implementation of this widely used algorithm in hardware in order to achieve very high performance by exploiting block pipelining and parallel processing. The two linear array architectures that we designed have been described as soft IP cores in fully parameterizable VHDL that can be used to synthesize effortlessly different k-NN parallel architectures for any desirable combination of the problem size parameters. They have been evaluated for a large variety of parameter combinations and Xilinx FPGAs. It is shown that they can be used to solve efficiently very large size k-NN classification problems, even with thousands of training vectors or vector dimensions, using a single, moderate size FPGA device. Furthermore the FPGA implementations exceed by a factor of two the performance of optimized NVIDIA CUDA API software implementations for the powerful GeForce 8800GTX GPU.

Protein spot detection and quantification in 2-DE gel images using machine-learning methods

Two‐dimensional gel electrophoresis (2‐DE) is the most established protein separation method used in expression proteomics. Despite the existence of sophisticated software tools, 2‐DE gel image analysis still remains a serious bottleneck. The low accuracies of commercial software packages and the extensive manual calibration that they often require for acceptable results show that we are far from achieving the goal of a fully automated and reliable, high‐throughput gel processing system. We present a novel spot detection and quantification methodology which draws heavily from unsupervised machine‐learning methods. Using the proposed hierarchical machine learning‐based segmentation methodology reduces both the number of faint spots missed (improves sensitivity) and the number of extraneous spots introduced (improves precision). The detection and quantification performance has been thoroughly evaluated and is shown to compare favorably (higher F‐measure) to a commercially available software package (PDQuest). The whole image analysis pipeline that we have developed is fully automated and can be used for high‐throughput proteomics analysis since it does not require any manual intervention for recalibration every time a new 2‐DE gel image is to be analyzed. Furthermore, it can be easily parallelized for high performance and also applied without any modification to prealigned group average gels.

Parallel architectures for the kNN classifier -- design of soft IP cores and FPGA implementations

We designed a variety of k-nearest-neighbor parallel architectures for FPGAs in the form of parameterizable soft IP cores. We show that they can be used to solve large classification problems with thousands of training vectors, or thousands of vector dimensions using a single FPGA, and achieve very high throughput. They can be used to flexibly synthesize architectures that also cover: 1NN classification (vector quantization), multishot queries (with different k), LOOCV cross-validation, and compare favorably to GPU implementations. To the best of our knowledge this is the first attempt to design flexible IP cores for the popular kNN classifier.

In silico modeling of the effects of alpha-synuclein oligomerization on dopaminergic neuronal homeostasis

BackgroundAlpha-synuclein (ASYN) is central in Parkinson’s disease (PD) pathogenesis. Converging pieces of evidence suggest that the levels of ASYN expression play a critical role in both familial and sporadic Parkinson’s disease. ASYN fibrils are the main component of inclusions called Lewy Bodies (LBs) which are found mainly in the surviving neurons of the substantia nigra. Despite the accumulated knowledge regarding the involvement of ASYN in molecular mechanisms underlying the development of PD, there is much information missing which prevents understanding the causes of the disease and how to stop its progression.ResultsUsing a Systems Biology approach, we develop a biomolecular reactions model that describes the intracellular ASYN dynamics in relation to overexpression, post-translational modification, oligomerization and degradation of the protein. Especially for the proteolysis of ASYN, the model takes into account the biological knowledge regarding the contribution of Chaperone Mediated Autophagy (CMA), macro-autophagic and proteasome pathways in the protein’s degradation. Importantly, inhibitory phenomena, caused by ASYN, concerning CMA (more specifically the lysosomal-associated membrane protein 2a, abbreviated as Lamp2a receptor, which is the rate limiting step of CMA) and the proteasome are carefully modeled. The model is validated by simulation studies of known experimental overexpression data from SH-SY5Y cells and the unknown model parameters are estimated either computationally or by experimental fitting. The calibrated model is then tested under three hypothetical intervention scenarios and in all cases predicts increased cell viability that agrees with experimental evidence. The biomodel has been annotated and is made available in SBML format.ConclusionsThe mathematical model presented here successfully simulates the dynamic phenomena of ASYN overexpression and oligomerization and predicts the biological system’s behavior in a number of scenarios not used for model calibration. It allows, for the first time, to qualitatively estimate the protein levels that are capable of deregulating proteolytic homeostasis. In addition, it can help form new hypotheses for intervention that could be tested experimentally.

A tool for simulation and geo-animation of wildfires with fuel editing and hotspot monitoring capabilities

FLogA (Fire Logic Animation) is a set of methods and an interactive, Web-based, user-friendly software tool which allows the user to draw a forest area on the map anywhere in Europe, insert fire ignition points, generate on the fly all input data layers required for a wildfire simulation, and then simulate and animate the behavior of the evolving fireline under different weather conditions. FLogA utilizes only publicly available non-proprietary data, software libraries and Web services. It adopts a distributed, open, service oriented architecture that is easy to maintain and extend. Wind, as the most dynamic parameter affecting a wildfires behavior, can be sampled around a reference direction and speed value reported by the closest METAR station, to generate multiple simulation scenarios. FLogA generates informative interactive geo-animations of simulation results with color representing a fire property of interest, such as the flame length or the forest cells burn probability, while the terrain of the forest in the background may be colored according to a characteristic of the forest (e.g. elevation, land cover, etc). Geo-animations allow the user to fly-over any part of the affected terrain as the fire is progressing. In addition, FLogA offers drawing tools for editing the forests spatial properties (e.g. change fuels, define cleanings zones etc.) to generate alternative what-if simulation scenarios. Furthermore, it can be set to automatically monitor any European forest area and trigger fire simulations as soon as hotspots are posted on the Internet by satellite services. Automatic generation of all input data for wildfire simulations all over Europe.Dynamic and interactive geo-animations of simulation output.Graphical editing of forest fuels to generate what-if simulation scenarios.Hotspot monitoring and automated wildfire simulation triggering for all Europe.Open, service oriented architecture using non-proprietary data layers and tools.

Flexible IP cores for the k-NN classification problem and their FPGA implementation

The k-nearest neighbor (k-NN) is a popular non-parametric benchmark classification algorithm to which new classifiers are usually compared. It is used in numerous applications, some of which may involve thousands of data vectors in a possibly very high dimensional feature space. For real-time classification a hardware implementation of the algorithm can deliver high performance gains by exploiting parallel processing and block pipelining. We present two different linear array architectures that have been described as soft parameterized IP cores in VHDL. The IP cores are used to synthesize and evaluate a variety of array architectures for a different k-NN problem instances and Xilinx FPGAs. It is shown that we can solve efficiently, using a medium size FPGA device, very large size classification problems, with thousands of reference data vectors or vector dimensions, while achieving very high throughput. To the best of our knowledge, this is the first effort to design flexible IP cores for the FPGA implementation of the widely used k-NN classifier.