Calibration of a Hall effect displacement measurement system for complex motion analysis using a neural network

Biomechanics studies often require the analysis of position and orientation. Although a variety of transducer and camera systems can be utilized, a common inexpensive alternative is the Hall effect sensor. Hall effect sensors have been used extensively for one-dimensional position analysis but their non-linear behavior and cross-talk effects make them difficult to calibrate for effective and accurate two- and three-dimensional position and orientation analysis. The aim of this study was to develop and calibrate a displacement measurement system for a hydraulic-actuation joystick used for repetitive motion analysis of heavy equipment operators. The system utilizes an array of four Hall effect sensors that are all active during any joystick movement. This built-in redundancy allows the calibration to utilize fully connected feed forward neural networks in conjunction with a Microscribe 3D digitizer. A fully connected feed forward neural network with one hidden layer containing five neurons was developed. Results indicate that the ability of the neural network to accurately predict the x, y and z coordinates of the joystick handle was good with r(2) values of 0.98 and higher. The calibration technique was found to be equally as accurate when used on data collected 5 days after the initial calibration, indicating the system is robust and stable enough to not require calibration every time the joystick is used. This calibration system allowed an infinite number of joystick orientations and positions to be found within the range of joystick motion.     

用空间直观模型是否足以从斑块占据性资料中推断集合种群的动态过程?

在集合种群的研究中,经常要根据空间占据性数据应用斑块模型来推断种群的动态过程,在保护生物学应用中,斑块占据性模型的参数估测对于阐释集合种群动态和预测种群对生境破坏的反应极为重要。我们探讨了一种广泛应用的空间直观模型——率函数模型(Incidence function model)中参数估测的不确定性问题,通过构建由50个斑块组成的网络和两个假想的已知参数的集合种群,应用模拟模型产生集合种群随时间变化的斑块占据性数据系列：即快照(snapshot)。然后,根据这些快照,应用率函数模型和最大似然法估测种群动态参数。此外,我们还给出了传统的率函数模型的一个变形,这个变形包含了目标区效应(Target area effect)：即一个斑块的占据概率不但取决于空间隔离度,也取决于斑块本身面积的大小。结果表明：根据同一个集合种群不同的快照所估测的参数可以有很大差异,一个快照得出的参数提示的是占据性强但存活率低的集合种群,而另一个快照可能反映的是一个占据性弱但存活率高的集合种群。应用传统的率函数模型于一个包含了目标区效应的集合种群,导致斑块大小相关的灭绝率参数估测的正偏差。因此,仅根据一个快照的空间占据性数据来推测集合种群的过程有很大的不确定性[动物学报49(6)：787～794,2003]。     

Neural network models for earthquake magnitude prediction using multiple seismicity indicators

Neural networks are investigated for predicting the magnitude of the largest seismic event in the following month based on the analysis of eight mathematically computed parameters known as seismicity indicators. The indicators are selected based on the Gutenberg-Richter and characteristic earthquake magnitude distribution and also on the conclusions drawn by recent earthquake prediction studies. Since there is no known established mathematical or even empirical relationship between these indicators and the location and magnitude of a succeeding earthquake in a particular time window, the problem is modeled using three different neural networks: a feed-forward Levenberg-Marquardt backpropagation (LMBP) neural network, a recurrent neural network, and a radial basis function (RBF) neural network. Prediction accuracies of the models are evaluated using four different statistical measures: the probability of detection, the false alarm ratio, the frequency bias, and the true skill score or R score. The models are trained and tested using data for two seismically different regions: Southern California and the San Francisco bay region. Overall the recurrent neural network model yields the best prediction accuracies compared with LMBP and RBF networks. While at the present earthquake prediction cannot be made with a high degree of certainty this research provides a scientific approach for evaluating the short-term seismic hazard potential of a region.     

Massively parallel implementations of theories for apparent motion

Two solutions for the correspondence problem for long-range motion are investigated. The first is a modification of the Minimal Mapping Theory (S. Ullman: The Interpretation of Visual Motion, MIT Press, Cambridge, 1979) that is implemented by a massively parallel network. In this network, every two units are interconnected, and thus, its convergence is fast and relatively independent of the number of image features. Computer simulations show that our method accounts as well as the Minimal Mapping Theory for apparent-motion phenomena, although some differences exist. Mathematical proofs provide conditions for the convergence of the network. The second 'solution' for the correspondence problem is called the Structural Theory. This theory assumes that the three-dimensional structure of viewed objects does not change fast in time. Then, the theory looks for the correspondence and three-dimensional structure that best fulfill this assumption. A massively parallel network implementation of this theory is also possible. However, its performance is poor due to the high complexity of its solution space. This supports Ullman's (1979) suggestion that the visual system separates the structure-from-motion process into two stages. First, a stage for motion measurement, and then a stage for structure recovery.     

Snapshot memories and landmark guidance in wood ants

Insects are thought to pinpoint a place by using memorized "snapshots," i.e., two-dimensional retinotopic views of the surrounding landmarks recorded when at the place (reviewed in ). Insects then reach the place by moving until their current view matches their snapshot. To determine when snapshots are recalled, and how differences between view and snapshot are translated into appropriate movements, we analyzed the approaches of wood ants to a feeding site that was located in the center of an array of two or three cylinders. In ants, contrary to flying hymenopterans, body orientation and direction of travel are collinear, so that an ant approaching an object always looks at it with frontal visual field. On their way to a food site, ants fixated and approached a cylinder predominantly when its angular size was smaller than when viewed from the food site. This finding implies that ants store snapshots at this place while fixating landmarks with frontal retina, so simplifying the later alignment of snapshots with their current view. It also means that ants recall snapshots well in advance of reaching the place. Although snapshots are centered on a landmark, we show that they extend at least 120 degrees into the periphery.     

Landmark maps for honeybees

Experiments by Fabre (1915), Thorpe (1950), Chmurzynski (1964), and most recently Gould (1986) suggest that insects have maps of their terrain which enable them to find their way directly to a goal when they are displaced several hundred metres from it. This paper discusses what might constitute an insect's map in terms of a two-part computational model. The first part describes how an insect reaches a goal when the insect is sufficiently close that it can see some of the landmarks which are visible from the goal. The second part considers the problem of navigating when there is no similarity between the view from the release-site and the view from the goal.We start from a model designed to explain how a bee might return to a goal using a two-dimensional snapshot of the landscape seen from the goal (Collett and Cartwright 1983). To guide its return, the model bee continuously compares its snapshot with its current retinal image and moves so as to reduce the discrepancy between the two. Bees can only be guided in the right direction by the difference between current retinal image and snapshot when there is some resemblance between the two. In a realistically cluttered world, snapshot and retinal image become very dis-similar only a short distance from the goal.To increase the distance from which a model bee can return, the bee takes two snapshots at the goal. The first snapshot excludes landmarks near to the goal and the second snapshot includes them. With close landmarks filtered from both snapshot and retinal image, the match between the two deteriorates gradually as the bee moves away from the goal. A model bee using a filtered snapshot and image finds its way back to the neighbourhood of the goal from a relatively long distance (Fig. 2). The bee then switches to the second snapshot and is guided to the precise spot by its memory of the close landmarks.For longer range guidance, the model bee is equipped with an album of snapshots, each taken at a different location within the terrain. Linked to each snapshot is a vector encoding the distance and direction from the place where the snapshot was taken to the hive. When the bee is displaced to a new position, it selects the snapshot which best matches its current image and follows the associated home-vector back to the hive (Fig. 3). Such a hive-centred map can also be used to devise novel routes to places other than the hive. For instance, a bee can reach a foraging site from anywhere in its terrain by adding the home-vector recalled at the starting position to a vector specifying the distance and direction of the foraging site from the hive. The sum of these two vectors defines a direct trajectory to the foraging site.     

Pre‐/post‐otic rhombomeric interactions control the emergence of a fetal‐like respiratory rhythm in the mouse embryo

How regional patterning of the neural tube in vertebrate embryos may influence the emergence and the function of neural networks remains elusive. We have begun to address this issue in the embryonic mouse hindbrain by studying rhythmogenic properties of different neural tube segments. We have isolated pre‐ and post‐otic hindbrain segments and spinal segments of the mouse neural tube, when they form at embryonic day (E) 9, and grafted them into the same positions in stage‐matched chick hosts. Three days after grafting, in vitro recordings of the activity in the cranial nerves exiting the grafts indicate that a high frequency (HF) rhythm (order: 10 bursts/min) is generated in post‐otic segments while more anterior pre‐otic and more posterior spinal territories generate a low frequency (LF) rhythm (order: 1 burst/min). Comparison with homo‐specific grafting of corresponding chick segments points to conservation in mouse and chick of the link between the patterning of activities and the axial origin of the hindbrain segment. This HF rhythm is reminiscent of the respiratory rhythm known to appear at E15 in mice. We also report on pre‐/post‐otic interactions. The pre‐otic rhombomere 5 prevents the emergence of the HF rhythm at E12. Although the nature of the interaction with r5 remains obscure, we propose that ontogeny of fetal‐like respiratory circuits relies on: (i) a selective developmental program enforcing HF rhythm generation, already set at E9 in post‐otic segments, and (ii) trans‐segmental interactions with pre‐otic territories that may control the time when this rhythm appears. © 2006 Wiley Periodicals, Inc. J Neurobiol, 2006     

Neural correlates of visual motion prediction

Predicting the trajectories of moving objects in our surroundings is important for many life scenarios, such as driving, walking, reaching, hunting and combat. We determined human subjects' performance and task-related brain activity in a motion trajectory prediction task. The task required spatial and motion working memory as well as the ability to extrapolate motion information in time to predict future object locations. We showed that the neural circuits associated with motion prediction included frontal, parietal and insular cortex, as well as the thalamus and the visual cortex. Interestingly, deactivation of many of these regions seemed to be more closely related to task performance. The differential activity during motion prediction vs. direct observation was also correlated with task performance. The neural networks involved in our visual motion prediction task are significantly different from those that underlie visual motion memory and imagery. Our results set the stage for the examination of the effects of deficiencies in these networks, such as those caused by aging and mental disorders, on visual motion prediction and its consequences on mobility related daily activities.     

A comparison of two non-invasive methods for measuring scapular orientation in functional positions

Identification of scapular dyskinesis and evaluation of interventions depend on the ability to properly measure scapulothoracic (ST) motion. The most widely used measurement approach is the acromion marker cluster (AMC), which can yield large errors in extreme humeral elevation and can be inaccurate in children and patient populations. Recently, an individualized regression approach has been proposed as an alternative to the AMC. This technique utilizes the relationship between ST orientation, humerothoracic orientation and acromion process position derived from calibration positions to predict dynamic ST orientations from humerothoracic and acromion process measures during motion. These individualized regressions demonstrated promising results for healthy adults; however, this method had not yet been compared to the more conventional AMC. This study compared ST orientation estimates by the AMC and regression approaches to static ST angles determined by surface markers placed on palpated landmarks in typically developing adolescents performing functional tasks. Both approaches produced errors within the range reported in the literature for skin-based scapular measurement techniques. The performance of the regression approach suffered when applied to positions outside of the range of motion in the set of calibration positions. The AMC significantly underestimated ST internal rotation across all positions and overestimated posterior tilt in some positions. Overall, root mean square errors for the regression approach were smaller than the AMC for every position across all axes of ST motion. Accordingly, we recommend the regression approach as a suitable technique for measuring ST kinematics in functional motion.     

Identification of Neural Networks That Contribute to Motion Sickness through Principal Components Analysis of Fos Labeling Induced by Galvanic Vestibular Stimulation

Motion sickness is a complex condition that includes both overt signs (e.g., vomiting) and more covert symptoms (e.g., anxiety and foreboding). The neural pathways that mediate these signs and symptoms are yet to identified. This study mapped the distribution of c-fos protein (Fos)-like immunoreactivity elicited during a galvanic vestibular stimulation paradigm that is known to induce motion sickness in felines. A principal components analysis was used to identify networks of neurons activated during this stimulus paradigm from functional correlations between Fos labeling in different nuclei. This analysis identified five principal components (neural networks) that accounted for greater than 95% of the variance in Fos labeling. Two of the components were correlated with the severity of motion sickness symptoms, and likely participated in generating the overt signs of the condition. One of these networks included neurons in locus coeruleus, medial, inferior and lateral vestibular nuclei, lateral nucleus tractus solitarius, medial parabrachial nucleus and periaqueductal gray. The second included neurons in the superior vestibular nucleus, precerebellar nuclei, periaqueductal gray, and parabrachial nuclei, with weaker associations of raphe nuclei. Three additional components (networks) were also identified that were not correlated with the severity of motion sickness symptoms. These networks likely mediated the covert aspects of motion sickness, such as affective components. The identification of five statistically independent component networks associated with the development of motion sickness provides an opportunity to consider, in network activation dimensions, the complex progression of signs and symptoms that are precipitated in provocative environments. Similar methodology can be used to parse the neural networks that mediate other complex responses to environmental stimuli.     

Orientation flights of solitary wasps (Cerceris; Sphecidae; Hymenoptera)

Bees and wasps are known to use a visual representation of the nest environment to guide the final approach to their nest. It is also known that they acquire this representation during an orientation flight performed on departure.A detailed film analysis shows that orientation flights in solitary wasps of the genus Cerceris consist of a systematic behavioural sequence: after lift-off from the nest entrance, wasps fly in ever increasing arcs around the nest. They fly along these arcs obliquely to their long axis and turn so that the nest entrance is held in the left or right visual field at retinal positions between 30° and 70° from the midline. Horizontal distance from the nest and height above ground increase throughout an orientation flight so that the nest is kept at retinal elevations between 45° and 60° below the horizon. The wasps' rate of turning is constant at between 100°/s and 200°/s independent of their distance from the nest and their ground velocity increases with distance. The consequence of this is that throughout the flight wasps circle at a constant angular velocity around the nest.Orientation flights are strongly influenced by landmark lay-out. Wasps adjust their flight-path and their orientation in a way that allows them to fixate the nest entrance and to hold the closest landmark in their frontal visual field.The orientation flight generates a specific topography of motion parallax across the visual field. This could be used by wasps to acquire a series of snapshots that all contain the nest position, to acquire snapshots of close landmarks only (distance filtering), to exclude shadow contours from their visual representation (figure-ground discrimination) or to gain information on the distance of landmarks relative to the nest.     

Sequential RAM-based neural networks: learnability, generalisation, knowledge extraction, and grammatical inference

A fundamental question in the field of artificial neural networks is what set of problems a given class of networks can perform (computability). Such a problem can be made less general, but no less important, by asking what these networks could learn by using a given training procedure (learnability). The basic purpose of this paper is to address the learnability problem. Specifically, it analyses the learnability of sequential RAM-based neural networks. The analytical tools used are those of Automata Theory. In this context, this paper establishes which class of problems and under what conditions such networks, together with their existing learning rules, can learn and generalize. This analysis also yields techniques for both extracting knowledge from and inserting knowledge into the networks. The results presented here, besides helping in a better understanding of the temporal behaviour of sequential RAM-based networks, could also provide useful insights for the integration of the symbolic/connectionist paradigms.     

Refinement of protein-protein complexes in contact map space with metadynamics simulations

Accurate protein-protein complex prediction, to atomic detail, is a challenging problem. For flexible docking cases, current state-of-the-art docking methods are limited in their ability to exhaustively search the high dimensionality of the problem space. In this study, to obtain more accurate models, an investigation into the local optimization of initial docked solutions is presented with respect to a reference crystal structure. We show how physics-based refinement of protein-protein complexes in contact map space (CMS), within a metadynamics protocol, can be performed. The method uses 5 times replicated 10 ns simulations for sampling and ranks the generated conformational snapshots with ZRANK to identify an ensemble of n snapshots for final model building. Furthermore, we investigated whether the reconstructed free energy surface (FES), or a combination of both FES and ZRANK, referred to as CS_α, can help to reduce snapshot ranking error.     

Location and connectivity of abdominal motoneurons in the embryo and larva of Drosophila melanogaster

The soma location and peripheral connectivity of motoneurons in abdominal segments of the embryo and larva of the fruitfly, Drosophila melanogaster are described as an initial step in determining the mechanisms by which motoneurons make connections with their target muscles in a genetically accessible organism. Embryonic motoneuron somata were retrogradely labelled by application of the fluorescent dye, DiI, to the whole peripheral nerve or to its separate anterior or posterior fascicles in segments A5-A7 of late stage 15/early stage 16 embryos. This technique reveals a stereotyped, segmentally repeated population of 34 motoneurons per hemisegment, several of which can be individually identified from their soma position. The same set of motoneurons was revealed in third instar larvae of D. melanogaster by cobalt backfilling of abdominal peripheral nerves, although the positions of some of these neurons change during larval development. The peripheral connectivity and axon morphology of several of the abdominal motoneurons was determined by intracellular injection with Lucifer Yellow in stage 16 embryos. For the motoneurons with axons in the anterior fascicle there is no clear relationship between somata groupings and the muscle targets innervated: contrary to earlier claims, these motoneurons arborize over both ventral and dorsal muscles. Individual motoneurons possess a stereotyped pattern of terminal arborization.     

Pre-/post-otic rhombomeric interactions control the emergence of a fetal-like respiratory rhythm in the mouse embryo

How regional patterning of the neural tube in vertebrate embryos may influence the emergence and the function of neural networks remains elusive. We have begun to address this issue in the embryonic mouse hindbrain by studying rhythmogenic properties of different neural tube segments. We have isolated pre- and post-otic hindbrain segments and spinal segments of the mouse neural tube, when they form at embryonic day (E) 9, and grafted them into the same positions in stage-matched chick hosts. Three days after grafting, in vitro recordings of the activity in the cranial nerves exiting the grafts indicate that a high frequency (HF) rhythm (order: 10 bursts/min) is generated in post-otic segments while more anterior pre-otic and more posterior spinal territories generate a low frequency (LF) rhythm (order: 1 burst/min). Comparison with homo-specific grafting of corresponding chick segments points to conservation in mouse and chick of the link between the patterning of activities and the axial origin of the hindbrain segment. This HF rhythm is reminiscent of the respiratory rhythm known to appear at E15 in mice. We also report on pre-/post-otic interactions. The pre-otic rhombomere 5 prevents the emergence of the HF rhythm at E12. Although the nature of the interaction with r5 remains obscure, we propose that ontogeny of fetal-like respiratory circuits relies on: (i) a selective developmental program enforcing HF rhythm generation, already set at E9 in post-otic segments, and (ii) trans-segmental interactions with pre-otic territories that may control the time when this rhythm appears.     

Recognition of distantly related protein sequences using conserved motifs and neural networks.

A sensitive technique for protein sequence motif recognition based on neural networks has been developed. It involves three major steps. (1) At each appropriate alignment position of a set of N matched sequences, a set of N aligned oligopeptides is specified with preselected window length. N neural nets are subsequently and successively trained on N-1 amino acid spans after eliminating each ith oligopeptide. A test for recognition of each of the ith spans is performed. The average neural net recognition over N such trials is used as a measure of conservation for the particular windowed region of the multiple alignment. This process is repeated for all possible spans of given length in the multiple alignment. (2) The M most conserved regions are regarded as motifs and the oligopeptides within each are used to train intensively M individual neural networks. (3) The M networks are then applied in a search for related primary structures in a databank of known protein sequences. The oligopeptide spans in the database sequence with strongest neural net output for each of the M networks are saved and then scored according to the output signals and the proper combination that follows the expected N- to C-terminal sequence order. The motifs from the database with highest similarity scores can then be used to retrain the M neural nets, which can be subsequently utilized for further searches in the databank, thus providing even greater sensitivity to recognize distant familial proteins. This technique was successfully applied to the integrase, DNA-polymerase and immunoglobulin families.     

Molecular docking of balanol to dynamics snapshots of protein kinase A

Even if the structure of a receptor has been determined experimentally, it may not be a conformation to which a ligand would bind when induced fit effects are significant. Molecular docking using such a receptor structure may thus fail to recognize a ligand to which the receptor can bind with reasonable affinity. Here, we examine one way to alleviate this problem by using an ensemble of receptor conformations generated from a molecular dynamics simulation for molecular docking. Two molecular dynamics simulations were conducted to generate snapshots for protein kinase A: one with the ligand bound, the other without. The ligand, balanol, was then docked to conformations of the receptors presented by these trajectories. The Lamarckian genetic algorithm in Autodock [Goodsell et al. J Mol Recognit 1996;9(1):1-5; Morris et al. J Comput Chem 1998;19(14):1639-1662] was used in the docking. Three ligand models were used: rigid, flexible, and flexible with torsional potentials. When the snapshots were taken from the molecular dynamics simulation of the protein-ligand complex, the correct docking structure could be recovered easily by the docking algorithm in all cases. This was an easier case for challenging the docking algorithm because, by using the structure of the protein in a protein-ligand complex, one essentially assumed that the protein already had a pocket to which the ligand can fit well. However, when the snapshots were taken from the ligand-free protein simulation, which is more useful for a practical application when the structure of the protein-ligand complex is not known, several clusters of structures were found. Of the 10 docking runs for each snapshot, at least one structure was close to the correctly docked structure when the flexible-ligand models were used. We found that a useful way to identify the correctly docked structure was to locate the structure that appeared most frequently as the lowest energy structure in the docking experiments to different snapshots.     

Neural network based temporal video segmentation

The organization of video information in video databases requires automatic temporal segmentation with minimal user interaction. As neural networks are capable of learning the characteristics of various video segments and clustering them accordingly, in this paper, a neural network based technique is developed to segment the video sequence into shots automatically and with a minimum number of user-defined parameters. We propose to employ growing neural gas (GNG) networks and integrate multiple frame difference features to efficiently detect shot boundaries in the video. Experimental results are presented to illustrate the good performance of the proposed scheme on real video sequences.     

6R instrumented spatial linkages for anatomical joint motion measurement--Part 2: Calibration.

The six-revolute-joint instrumented spatial linkage (6R ISL) is often the measurement system of choice for monitoring motion of anatomical joints. However, due to tolerances of the linkage parameters, the system may not be as accurate as desired. A calibration algorithm and associated calibration device have been developed to refine the initial measurements of the ISL's mechanical and electrical parameters so that the measurement of six-degree-of-freedom motion will be most accurate within the workspace of the anatomical joint. The algorithm adjusts the magnitudes of selected linkage parameters to reduce the squared differences between the six known and calculated anatomical position parameters at all the calibration positions. Weighting is permitted so as to obtain a linkage parameter set that is specialized for measuring certain anatomical position parameters. Output of the algorithm includes estimates of the measuring system accuracy. For a particular knee-motion-measuring ISL and calibration device, several interdependent design parameter relationships have been identified. These interdependent relationships are due to the configuration of the ISL and calibration device, the number of calibration positions, and the limited resolution of the devices that monitor the position of the linkage joints. It is shown that if interdependence is not eliminated, then the resulting ISL parameter set will not be accurate in measuring motion outside of the calibration positions, even though these positions are within the ISL workspace.