Spatial Pattern Recognition Research Articles

Spatial defect pattern recognition on semiconductor wafers using model-based clustering and Bayesian inference

Defects on semiconductor wafers tend to cluster and the spatial defect patterns contain useful information about potential problems in the manufacturing process. This study proposes to use model-based clustering algorithms via Bayesian inferences for spatial defect pattern recognition on semiconductor wafers. These new algorithms can find the number of defect clusters as well as identify the pattern of each cluster automatically. They are capable of detecting curvilinear patterns, ellipsoidal patterns and nonuniform global defect patterns. Promising results have been obtained from simulation studies.

Automatic recognition and statistical quantification of spatial patterns of gene expression in zebra finch brain in response to auditory stimulation

Introduction We present here an automated procedure for detecting and mapping spatially neuronal cells stained by in situ hybridization. In songbird brain, the subcellular localization of song-induced activity-dependent mRNA is timedependent [2]. Thus it is possible to distinguish neuronal cells activated at different time points and therefore the representation of two different stimuli within the same animal. Images captured through a fluorescent microscope are analyzed off-line individually and then stitched together. Furthermore, neighboring brain slices can be stacked together, resulting in a 3D map of the spatial pattern of gene expression in any given brain structure. from Seventeenth Annual Computational Neuroscience Meeting: CNS*2008 Portland, OR, USA. 19–24 July 2008

A new approach to the spatio-temporal pattern identification in neuronal multi-electrode registrations

A lot of methods were created in last decade for the spatio-temporal analysis of multi-electrode array (MEA) neuronal data sets. All these methods were implemented starting from a channel to channel analysis, with a great computational effort and onerous spatial pattern recognition task. Our idea is to approach the MEA data collection from a different point of view, i.e. considering all channels simultaneously. We transform the 2D plus time MEA signal in a mono-dimensional plus time signal and elaborate it as a normal 1D signal, using the Space-Amplitude Transform method. This geometrical transformation is completely invertible and allows to employ very fast processing algorithms.

A new approach to the spatio-temporal pattern identification in neuronal multi-electrode registrations

AbstractA lot of methods were created in last decade for the spatio-temporal analysis of multi-electrode array (MEA) neuronal data sets. All these methods were implemented starting from a channel to channel analysis, with a great computational effort and onerous spatial pattern recognition task. Our idea is to approach the MEA data collection from a different point of view, i.e. considering all channels simultaneously. We transform the 2D plus time MEA signal in a mono-dimensional plus time signal and elaborate it as a normal 1D signal, using the Space-Amplitude Transform method. This geometrical transformation is completely invertible and allows to employ very fast processing algorithms.

Spatial chemical conservation of hot spot interactions in protein-protein complexes

BackgroundConservation of the spatial binding organizations at the level of physico-chemical interactions is important for the formation and stability of protein-protein complexes as well as protein and drug design. Due to the lack of computational tools for recognition of spatial patterns of interactions shared by a set of protein-protein complexes, the conservation of such interactions has not been addressed previously.ResultsWe performed extensive spatial comparisons of physico-chemical interactions common to different types of protein-protein complexes. We observed that 80% of these interactions correspond to known hot spots. Moreover, we show that spatially conserved interactions allow prediction of hot spots with a success rate higher than obtained by methods based on sequence or backbone similarity. Detection of spatially conserved interaction patterns was performed by our novel MAPPIS algorithm. MAPPIS performs multiple alignments of the physico-chemical interactions and the binding properties in three dimensional space. It is independent of the overall similarity in the protein sequences, folds or amino acid identities. We present examples of interactions shared between complexes of colicins with immunity proteins, serine proteases with inhibitors and T-cell receptors with superantigens. We unravel previously overlooked similarities, such as the interactions shared by the structurally different RNase-inhibitor families.ConclusionThe key contribution of MAPPIS is in discovering the 3D patterns of physico-chemical interactions. The detected patterns describe the conserved binding organizations that involve energetically important hot spot residues and are crucial for the protein-protein associations.

Defect spatial pattern recognition using a hybrid SOM–SVM approach in semiconductor manufacturing

As manufacturing geometries continue to shrink and circuit performance increases, fast fault detection and semiconductor yield improvement is of increasing concern. Circuits must be controlled to reduce parametric yield loss, and the resulting circuits tested to guarantee that they meet specifications. In this paper, a hybrid approach that integrates the Self-Organizing Map and Support Vector Machine for wafer bin map classification is proposed. The log odds ratio test is employed as a spatial clustering measurement preprocessor to distinguish between the systematic and random wafer bin map distribution. After the smoothing step is performed on the wafer bin map, features such as co-occurrence matrix and moment invariants are extracted. The wafer bin maps are then clustered with the Self-Organizing Map using the aforementioned features. The Support Vector Machine is then applied to classify the wafer bin maps to identify the manufacturing defects. The proposed method can transform a large number of wafer bin maps into a small group of specific failure patterns and thus shorten the time and scope for troubleshooting to yield improvement. Real data on over 3000 wafers were applied to the proposed approach. The experimental results show that our approach can obtain over 90% classification accuracy and outperform back-propagation neural network.

Spatial pattern recognition of extreme temperature climatology: assessing HadCM3 simulations via NCEP reanalyses over Europe

This article attempts to quantify the spatial uncertainties associated with extreme temperature’s response, by assessing data derived from climate model. This is undertaken by a comparison of the spatial pattern of a long-term time-series aggregation (1960/61-1989/90) for extreme temperatures simulated by a particular GCM (the UK Met Office - Hadley Centre climate model, HadCM3) to that of the USA NCAR NCEP Reanalyses, which are considered as ‘truth’, over the MICE (Modelling the Impacts of Climate Extremes - EU Project) spatial domain. Since evaluation of models is crucial to assessing future scenarios, the aim of this study is to investigate whether the extreme values predicted by the HadCM3 climate model can simulate those produced by NCEP Reanalyses, assuming that the extremes of both models are realizations of the same spatial stochastic process. To get more useful information about the uncertainties surrounding spatial climate projection, one also has to analyze the pattern of temperature extremes in terms of their anomalies. A common technical issue in the assessment of numerical spatial models is based on the Principal Components Analysis and Bayesian Classification for spatial pattern recognition. These methodologies are very important and useful for guiding an evolutionary statistical model-building process. This study leads to the conclusion that the HadCM3 Simulations do not realistically reproduce the NCEP Reanalyses, despite the fact that the climatology of extremes has demonstrated very similar spatial patterns. It is likely therefore that such instability may persist in the future.

Discrimination among similar looking noisy color patches using Margin Setting

The goal of discrimination of one color from many other similar-appearing colors even when the colored objects show substantial variation or noise is of obvious import. We show how to accomplish that using a technique called Margin Setting. It is possible not only to have very low error rates but also to have some control over the types of errors that do occur. Robust spectral filtering prior to spatial pattern recognition allows subsequent filtering processes to be based on conventional coherent optical correlation that can be done monochromatically.

Recognition of defect spatial patterns in semiconductor fabrication

This research develops an analytical structure comprised of a filtering scheme and a fuzzy rule-based inference system to help identify defect spatial patterns. These defect spatial patterns include ring, scratch, zone and repeating types. A set of image processing masks is designed to locate defect positions and then the filtering scheme is applied to extract defect clusters on wafers. With clearly identified defect clusters, three features of minimum rectangle area are used to locate and describe the shape of defect clusters. When all possible defect clusters are well represented by minimum rectangle areas, a set of fuzzy rules are established to combine all the defect clusters and therefore defect spatial patterns are identified. The experiments show that the present approach can effectively identify different defect spatial patterns on wafers.

Fish Perform Spatial Pattern Recognition and Abstraction by Exclusive Use of Active Electrolocation

The field generated by the electric organ of weakly electric fish varies with the electrical properties of nearby objects [1]. Correspondingly, current fluxes in this field differentially stimulate the electroreceptors in the fish's skin [2]. Thus, resistors are to conductors and insulators as gray is to black and white in optics. Additionally, the capacitances of plants and insect larvae contrast with those of water or stones, giving effects comparable to “coloration” [3]. Receptors [4] arrayed over a large area of the skin act like a retina upon which the discharge projects “electric images” [5]. By further central processing, the fish also discriminate between objects according to their composition [6], size, or distance [7], a procedure termed “electrolocation” [8], analogous to echolocation in bats [9]. Here we demonstrate that G. petersii and S. macrurus can also recognize 3D orientations and configurations and extract and generalize spatial features solely with their electrical sense. We presented fish with virtual electrical “objects” formed from electrodes set flush in the inner surface of a Y maze with various patterns of external connectivity. With reward and aversion training, the fish could recognize similar electrode configurations and extract a feature, e.g., a vertical connectivity, present in various novel configurations. Previously, shape recognition has only been shown in electrolocating fish when they are in full mechanical contact with solid objects [10].

Spatial heterogeneity in the herbaceous layer of a semi-arid savanna ecosystem

Despite increasing recognition of the role spatial pattern can play in ecosystem function, few studies have quantified spatial heterogeneity in savanna ecosystems. The spatial distribution of herbaceous biomass and species composition was measured across three scales in a semi-arid savanna in central Kenya, and patterns were related to environmental variables at different scales. Herbaceous biomass declined across a rainfall gradient and from upper to lower topographic positions, but variation within a site (across 5‐50 m) was similar in magnitude to among-site variation associated with rainfall and topography. Geostatistical analyses showed that patchiness at scales of 5‐25 m explained 20% of total variation in herbaceous biomass. This pattern arose from the presence of both 5‐10-m diameter patches containing high herbaceous biomass (> 170 g m 2 ) and 5‐10-m diameter patches characterized by nearly bare soil surfaces (< 40 g m 2 ). Patch structure was contingent on topography, with larger bare patches at ridgeline and upper hillslope positions. Grass species distributions showed the greatest degree of patch structure and species turnover across distances of 5‐45 m. Additional community variation was associated with topography, with minimal variation in species composition across the rainfall gradient. Pattern diversity significantly exceeded levels reported for four other grassland ecosystems, suggesting fundamental differences in local processes generating spatial pattern. It is hypothesized that heterogeneously distributed grazing pressure, interacting with the distribution of shrub canopies, is an important factor generating such high levels of small-scale patch structure in this savanna.

Statistical methods for visual defect metrology

Automated systems are used to inspect unpatterned and product wafers for particulates and other defects. Wafer defect count and defect density statistics are used as process control parameters, but are known to be deceptive in the presence of defect clustering. An improvement path using novel visual defect metrology statistical analyses is proposed. Quadrat analysis, nested analysis of variance, and principal component analysis use data available currently. Spatial point pattern statistics and spatial pattern recognition require special algorithms. Future process control systems made possible by these statistical analyses are discussed.

Statistical methods for visual defect metrology

Automated systems are used to inspect unpatterned and product wafers for particulates and other defects. Wafer defect count and defect density statistics are used as process control parameters, but are known to be deceptive in the presence of defect clustering. An improvement path using novel visual defect metrology statistical analyses is proposed. Quadrat analysis, nested analysis of variance, and principal component analysis use data available currently. Spatial point pattern statistics and spatial pattern recognition require special algorithms. Future process control systems made possible by these statistical analyses are discussed.

Adaptive Pattern Recognition

A long term goal of research in artificial intelligence is to determine and to implement principles which permit a movable machine to direct its actions depending upon sensory feed-back from its environment. This paper concentrates on spatial sensors which input images (2-dimensional arrays). A proposal is put forward in which the machine adapts to the actual data, and examples are given of input prediction, of detection of unexpected events, and of recognition of spatial patterns. The image sequence is locally partitioned into temporally contiguous subsequences of a fixed spatial extent. The spatial extent is constant over time and the temporal extent of a subsequence is maximized subject to the condition that the subsequence has occurred previously. The principle is illustrated on image sequences. It is further demonstrated on images which are structured as pseudo-temporal sequences of their rows. The demonstrations use diverse complex and simple examples to illustrate the versatility of the method. The demonstrations show that to a large degree it is not necessary for the user to supply explicit models for different pattern recognition tasks.

Memory, attention, and executive function in chronic fatigue syndrome.

To examine cognitive function in chronic fatigue syndrome. Twenty patients with chronic fatigue syndrome recruited from primary care and 20 matched normal controls were given CANTAB computerised tests of visuospatial memory, attention, and executive function, and verbal tests of letter and category fluency and word association learning. Patients with chronic fatigue syndrome were impaired, predominantly in the domain of memory but their pattern of performance was unlike that of patients with amnesic syndrome or dementia. They were normal on tests of spatial pattern recognition memory, simultaneous and delayed matching to sample, and pattern-location association learning. They were impaired on tests of spatial span, spatial working memory, and a selective reminding condition of the pattern-location association learning test. An executive test of planning was normal. In an attentional test, eight subjects with chronic fatigue syndrome were unable to learn a response set; the remainder exhibited no impairment in the executive set shifting phase of the test. Patients with chronic fatigue syndrome were also impaired on verbal tests of unrelated word association learning and letter fluency. Patients with chronic fatigue syndrome have reduced attentional capacity resulting in impaired performance on effortful tasks requiring planned or self ordered generation of responses from memory.

Associative Memory Based on Nonlinear Optical Medium Properties

Abstract Nonlinear-medium properties are examined for optical implementation of an associative memory system. The approach based on these properties allows one to consider two cases of the optical associative memory: spatial pattern recognition and recognition of spectral distributions. The model proposed for this purpose permits nonlinearity in the correlation domain to be used to reduce cross-correlation noise.

Spatial pattern recognition in a high energy particle detector using a neural network algorithm

An intrinsically parallel algorithm based on a Hopfield style neural network is used to associate the energy depositions in a multi-component, non-magnetic high energy particle detector. The algorithm has been implemented on a vector pipeline computer with greatly enhanced performance.

How birds identify species information in song: A pattern recognition approach

Pattern recognition theory (PRT) is a diverse array of models concerned with the recognition of spatial and temporal patterns by humans and machines. We have sought to use the theory to help us understand how passerines identify species information in conspecific song. In this paper, application of PRT provided (a) a tentative model for song recognition, (b) an improved methodology for song playback experiments, and (c) a theoretical analysis of the representation of the temporal structure of song. In each instance, PRT suggested better experiments and more detailed analyses than had been available previously.

Conceptual Framework for the Analysis and Classification of Fossil Concentrations

Densely fossiliferous deposits are receiving increasing attention for their yield of paleobiologic data and their usefulness in sedimentology and stratigraphy. This trend has created a pressing need for standardized descriptive terminology and a genetic classification based on a coherent conceptual framework. The descriptive procedure outlined here for skeletal concentrations stresses four features -taxonomic composition, bioclastic fabric, geometry, and internal structure-that can be described readily in the field by nonspecialists. The genetic classification scheme is based on three end members, representing biologic, sedimentologic, and diagenetic factors in skeletal concentration. Concentrations created through the simultaneous or sequential action of two or more factors are classified as mixed types. As a conceptual framework for comparative biostratinomic analysis, the broad categories of this ternary classification scheme should facilitate recognition of large-scale temporal and spatial patterns in skeletal accumulation. The usefulness of this approach is suggested by the good agreement between biostratinomic patterns observed in ancient onshore-offshore facies tracts and those predicted across paleobathymetric transects based on modern processes of skeletal concentration.

Neural representation of complex spatial patterns in peripheral afferents and somatosensory cortex of the awake monkey

The human capacity for tactual, spatial pattern recognition has been investigated by a number of people using embossed letters and Braille characters and the Optacon. The experiments reported here use identical stimuli in neurophysiological experiments to study the neural mechanisms underlying this spatial pattern recognition. The approach is to reconstruct the spatiotemporal patterns of neural activity evoked by these stimuli in the periphery and the cortex. The aim is to determine which pathways convey the critical information underlying pattern recognition and to analyze the transformations as the information proceeds centrally. The results of these experiments are “neural images” which we believe represent the distributed patterns of neural activity as the brain sees them. As expected, these images are mildly distorted replicas of the initial stimuli at the level of the periphery, i.e., the letter A evokes distributed patterns of neural activity in the form of an A. At the level of the cortex the “neu...