Showing papers on "Centroid published in 1995"

Facial feature extraction method and apparatus for a neural network acoustic and visual speech recognition system

Abstract: A facial feature extraction method and apparatus uses the variation in light intensity (gray-scale) of a frontal view of a speaker's face. The sequence of video images are sampled and quantized into a regular array of 150×150 pixels that naturally form a coordinate system of scan lines and pixel position along a scan line. Left and right eye areas and a mouth are located by thresholding the pixel gray-scale and finding the centroids of the three areas. The line segment joining the eye area centroids is bisected at right angle to form an axis of symmetry. A straight line through the centroid of the mouth area that is at right angle to the axis of symmetry constitutes the mouth line. Pixels along the mouth line and the axis of symmetry in the vicinity of the mouth area form a horizontal and vertical gray-scale profile, respectively. The profiles could be used as feature vectors but it is more efficient to select peaks and valleys (maximas and minimas) of the profile that correspond to the important physiological speech features such as lower and upper lip, mouth corner, and mouth area positions and pixel values and their time derivatives as visual vector components. Time derivatives are estimated by pixel position and value changes between video image frames. A speech recognition system uses the visual feature vector in combination with a concomitant acoustic vector as inputs to a time-delay neural network.

Describing motion for recognition

Abstract: Our goal is to describe motion of a moving human figure in order to recognize individuals by variation in the characteristics of the motion description. We begin with a short sequence of images of a moving figure, taken by a static camera, and derive dense optical flow data for the sequence. We determine a range of scale-independent features of each how image as a whole, ranging from the motion of the centroid of the moving points (assuming a static background), to the integral of the torque relative to the centroid. We then analyze the periodic structure of these sequences. All elements are multiples of the fundamental period of the gait, but they differ in phase. The phase is time-invariant, since it is independent of the sampling period. We show that there are several regularities in the phase differences of the signals. Moreover, some scalar measures of the signals may be useful in recognition. The representation is model-free, and therefore could be used to characterize the motion of other non-rigid bodies.

Roles for Spectral Centroid and Other Factors in Determining "Blended" Instrument Pairings in Orchestration

Abstract: Three perceptual experiments using natural-sounding instrument tones arranged in concurrently sounding pairs investigate a problem of orchestration: what factors determine selection of instruments to achieve various degrees of "blend" (fusion of multiple timbres into a single timbral image). The principal finding concerns the spectral centroid of the instruments (the midpoint of the spectral energy distribution). Blend worsened as a function of the overall centroid height of the combination (the centroid of the composite spectrum of the pair) or as the amount of difference between the centroids of the two instruments increased. Slightly different results were found depending on whether the instruments were on the same pitch or separated by a minor third. For unisons, composite centroid, attack similarity, and loudness envelope correlation accounted for 51% of the variance of blend. For minor thirds, centroid difference, composite centroid, attack similarity, and synchrony of offset accounted for 63% of the variance of blend. In a third experiment, instruments were manipulated to have different centroid levels to test if centroid made an independent contribution to blend. The results show that changes in centroid affect blend even when that is the only aspect of the sound that is changing. The findings create the potential for an approach to orchestration based on abstract properties of sound as a substitute for the traditional approach of teaching entirely by example.

A variational centroid density procedure for the calculation of transmission coefficients for asymmetric barriers at low temperature

Abstract: The low temperature behavior of the centroid density method of Voth, Chandler, and Miller (VCM) [J. Chem. Phys. 91, 7749 (1989)] is investigated for tunneling through a one‐dimensional barrier. We find that the bottleneck for a quantum activated process as defined by VCM does not correspond to the classical bottleneck for the case of an asymmetric barrier. If the centroid density is constrained to be at the classical bottleneck for an asymmetric barrier, the centroid density method can give transmission coefficients that are too large by as much as five orders of magnitude. We follow a variational procedure, as suggested by VCM, whereby the best transmission coefficient is found by varying the position of the centroid until the minimum value for this transmission coefficient is obtained. This is a procedure that is readily generalizable to multidimensional systems. We present calculations on several test systems which show that this variational procedure greatly enhances the accuracy of the centroid density method compared to when the centroid is constrained to be at the barrier top. Furthermore, the relation of this procedure to the low temperature periodic orbit or ‘‘instanton’’ approach is discussed.

Precision tracking based on segmentation with optimal layering for imaging sensors

Abstract: In the authors' previous work Oron, Kumar, and Bar-Shalom (1993), they presented a method for precision tracking of a low observable target based on data obtained from imaging sensors. The image was divided into several layers of gray level intensities and thresholded. A binary image was obtained and grouped into clusters using image segmentation techniques. Using the centroid measurements of the clusters, the probabilistic data association filter (PDAF) was employed for tracking the target centroid. In this correspondence, the division of the image into several layers of gray level intensities is optimized by minimizing the Bayes risk. This optimal layering of the image has the following properties: (1) following the segmentation, a closed-form analytical expression is obtained for the noise variance of the centroid measurement based on a single frame; (2) in comparison to the previous paper, the measurement noise variance is smaller by at least a factor of 2, thus improving the performance of the tracker. The usefulness of the method for practical applications is demonstrated by considering a sequence of real target images (a moving car) of about 20 pixels in size in a noisy urban environment where the measurement noise was calculated as having 0.32 pixel RMS value. Filtering with the PDAF further reduces this by a factor of 1.6. >

The Centroid of Points with Approximate Weights

Abstract: Let S be a set of points in ℝd, each with a weight that is not known precisely, only known to fall within some range What is the locus of the centroid of S? We prove that this locus is a convex polytope, the projection of a zonotope in ℝd+1 We derive complexity bounds and algorithms for the construction of these “centroid polytopes”

Adaptive optics and the outer scale of turbulence

Abstract: The outer-scale influence on spatial and temporal characteristics of turbulence-induced wave-front distortions is discussed. The calculation results for the image centroid and the Zernike modes are presented. Two methods of the outer-scale estimation based on tilt and defocus and on image centroid measurements are suggested. The application of the results obtained to adaptive optics problems is considered. Finally, a possibility of prediction of the wave-front statistical characteristics is discussed.

Recognizing 3D objects using photometric invariant

Abstract: We describe an efficient algorithm for recognizing 3D objects by combining photometric, and geometric invariants. A photometric property is derived, that is invariant to the changes of illumination and to relative object motion with respect to the camera and/or the lighting source in 3D space. We argue that conventional color constancy algorithms can not be used in the recognition of 3D objects. Further we show recognition does not require a full constancy of colors, rather, it only needs something that remains unchanged under the varying light conditions and poses of the objects. Combining the derived color invariant and the spatial constraints on the object surfaces, we identify corresponding positions in the model and the data space coordinates, using centroid invariance of corresponding groups of feature positions. Tests are given to show the stability and efficiency of our approach to 3D object recognition. >

Training a neural network using centroid dithering by randomly displacing a template

Abstract: Pattern recognition, for instance optical character recognition, is achieved by training a neural network, scanning an image, segmenting the image to detect a pattern, preprocessing the detected pattern, and applying the preprocessed detected pattern to the trained neural network. The preprocessing includes determining a centroid of the pattern and centrally positioning the centroid in a frame containing the pattern. The training of the neural network includes randomly displacing template patterns within frames before applying the template patterns to the neural network.

Mobile robot localization by remote viewing of a colored cylinder

Abstract: To visually determine the position and orientation of a mobile robot from a fixed location in its vicinity, the authors have employed a cylindrical target which has different colors in each of its four quadrants. By judicious selection of the colors, segmentation of imagery from the fixed location can determine the size and centroid of the cylinder, as well as the visible color quadrants. Both the cylinder size in monocular images, and the centroid disparity in stereo pairs, are shown to provide a measure of distance. The angle of the cylinder is determined by analyzing which color quadrants are visible and to what degree. Implementation and experimental testing of this technique shows that it provides accurate localization data to within one or two pixels of error.

Feature selection for object tracking in traffic scenes

Abstract: This paper describes a motion-analysis system, applied to the problem of vehicle tracking in real-world highway scenes. The system is structured in two stages. In the first one, a motion- detection algorithm performs a figure/ground segmentation, providing binary masks of the moving objects. In the second stage, vehicles are tracked for the rest of the sequence, by using Kalman filters on two state vectors, which represent each target's position and velocity. A vehicle's motion is represented by an affine model, taking into account translations and scale changes. Three types of features have been used for the vehicle's description state vectors. Two of them are contour-based: the bounding box and the centroid of the convex polygon approximating the vehicles contour. The third one is region-based and consists of the 2-D pattern of the vehicle in the image. For each of these features, the performance of the tracking algorithm has been tested, in terms of the position error, stability of the estimated motion parameters, trace of the motion model's covariance matrix, as well as computing time. A comparison of these results appears in favor of the use of the bounding box features.

Synthesis of Trumpet Tones Using a Wavetable and a Dynamic Filter

Abstract: The spectral response of the trumpet varies dramatically depending on its performed pitch and dynamic. The shape of a spectrum can be measured in terms of its spectral centroid, which is closely related to subjective brightness. We observe that trumpet spectra with the same centroid have similar spectral shapes regardless of their pitches. Based on a training set of 15 trumpet tones, a family of spectral envelopes has been created, each formed by averaging spectra with similar centroid values. For synthesis, a source spectrum is first sampled from the highest centroid spectral envelope. The remaining spectral envelopes are approximated by means of a variable low-pass filter. For each note produced, the filter is swept according to a time-varying centroid. Amplitude and frequency control functions are also matched. The result is a method that efficiently synthesizes convincing trumpet tones across the instrument's pitch and dynamic ranges using a wavetable and a centroid-controlled filter.

Method and device for positioning circuit board inspection device

Abstract: PURPOSE: To precisely position at high speed with a single compensation by rotating a fixture fixing a probe pin group CONSTITUTION: A circuit board 54, acting as an inspection object and fixed on an inspection table 11, is, based on an instruction from a computer 23, moved by an X/Y table, so that image recognizing two points patterns 55 and 56 of the substrate 54 are introduced into a field of view of a CCD camera 16 The patterns 55 and 56 are made into pictures by the camera 16, and converted into a binary image by an image processor device 22, and respective centroid coordinate positions 43 and 44 are stored (23) Then the table 20 is moved by deviation amount between a middle point 45, of the coordinates 43 and 44, and a middle point 48, of two reference position coordinates 46 and 47 set in advance, so as to coincide the points 45 and 48 Then, by the angle formed by both straight lines linking the coordinates 43 and 44, and the coordinates 46 and 47 respectively, a fixture consisting of a probe pin group 14 and a holding member 15 is rotated (17), for positioning a probe point Thus the coordinate of the table 11 does not vary, with no feedback control required, so that precise positioning is enabled at high speed

Surface sampling techniques for 3D object inspection

Abstract: While the uniform sampling method is quite popular for pointwise measurement of manufactured parts, this paper proposes three novel sampling strategies which emphasize 3D non-uniform inspection capability. They are: (a) the adaptive sampling, (b) the local adjustment sampling, and (c) the finite element centroid sampling techniques. The adaptive sampling strategy is based on a recursive surface subdivision process. Two different approaches are described for this adaptive sampling strategy. One uses triangle patches while the other uses rectangle patches. Several real world objects were tested using these two algorithms. Preliminary results show that sample points are distributed more closely around edges, corners, and vertices as desired for many classes of objects. Adaptive sampling using triangle patches is shown to generally perform better than both uniform and adaptive sampling using rectangle patches. The local adjustment sampling strategy uses a set of predefined starting points and then finds the local optimum position of each nodal point. This method approximates the object by moving the points toward object edges and corners. In a hybrid approach, uniform points sets and non-uniform points sets, first preprocessed by the adaptive sampling algorithm on a real world object were then tested using the local adjustment sampling method. The results show that the initial point sets when preprocessed by adaptive sampling using triangle patches, are moved the least amount of distance by the subsequently applied local adjustment method, again showing the superiority of this method. The finite element sampling technique samples the centroids of the surface triangle meshes produced from the finite element method. The performance of this algorithm was compared to that of the adaptive sampling using triangular patches. The adaptive sampling with triangular patches was once again shown to be better on different classes of objects.

Centroid oscillation gauge

Abstract: PURPOSE:To provide a centroid oscillation gage by which a degree of oscillation can be accurately found from many found centroid positions and furthermore, obtained degree of oscillation can be objectively and easily judged CONSTITUTION:This gage is equipped with a detection plate 11 on which each leg of an examinee is placed, plural load detecting means 12 attached on the detection plate 11 to detect the center of entire load applied on each leg of the examinee, an arithmetic means 13 which calculates the centroid position of the examinee based on a detection signal from the load detecting means 12, an XY coordinate position storage means 14 which converts the centroid position calculated by the arithmetic means 13 to a position set on X-Y coordinates in advance and also, stores converted continuous or intermittent XY coordinate positions as many XY coordinates, and an effective value calculation means 16 which finds an effective value RMS which shows distance from those coordinates (A, B) by stored XY coordinate positions and X coordinate A, Y coordinate B set in advance

A CMOS current-mode design of modified learning-vector-quantization neural networks

Abstract: A modification of LVQ model, Modified LVQ (MLVQ) model, is proposed for the estimation of centroid in pattern recognition. Computer simulation results are presented which demonstrate the behavior of the MLVQ model in estimating the class centroid by utilizing the distance-dependent step size. The results indicate the high potential of less dependence on the initial point as well as the precise settlement of the weight vectors to the centroids. The main feature is that the proposed model is robust to the noise perturbation between two pattern distributions in practical applications.

A new learning scheme for the recognition of dynamical handwritten characters

Abstract: Vector comparison is essential in pattern recognition. Numerous methods based on distance computation are available to carry out such comparison. Unfortunately most of them are applicable only if the vectors are of the same length or do not take into account components misalignment. This paper presents a new distance between two representations called the elastic distance and based on the dynamic programming technique. Properties are studied. We show that it leads to a variant of the least vector quantisation technique that learns the best representants of a group of prototypes. A new centroid computation algorithm is proposed. Finally, the learning scheme algorithm has been successfully applied on an online numerical handwritten character recognition problem using a previously computed centroid of a set of prototypes.

Image processing method

Abstract: PURPOSE:To reduce processing cost and to improve stability by determining the centroid location of an object by performing the evaluation of localization and recognizing an attribute with the obtained centroid location as a center. CONSTITUTION:An analog/digital conversion is performed for the image by the analog signal surrounded by a camera, etc., in an image input part 1 and the image becomes the original image 5 by a digital signal. This original image 5 is delivered to a preprocessing part 2, a preprocessing is performed for the image, the image becomes the preprocessed image where high frequency noise and the slow luminance change in the image are removed, and the image is inputted in an object extraction part 3 and an object recognition part 4. In the object extraction part 3, the localization of the inputted preprocessed image 6 is remarked, the centroid location 7 of an object is extracted and it is delivered to the object recognition part 4. In the object recognition part 4, the attribute 8 of the object is extracted from the centroid location 7 of this object and the preprocessed image 6 inputted from the preprocessing part 2 and the attribute is outputted to the outside. When the object extraction part 3 extracts plural objects, the attribute 8 can be recognized in parallel in the object recognition part 4.

Three-dimensional object shape recognition using cross sections

Abstract: This paper describes a computationally efficient 3D object surface matching algorithm. In the proposed method, object and model surfaces are scaled to be in a unit cube in the 3D space. They are then sliced along the magnitude axis and the resultant object and model surface cross sections are represented in binary image format. The cross- sections' centroids of an unknown object and the models of different shapes are computed in their respective binary images. The resultant cross-sections are translated to the origin of the spatial plane using the centroids. Major and minor axes of the plane cross sections are aligned with the coordinate axes of the spatial plane. Matching of the aligned cross sections is done in the direction of the gradient of the cross section boundary by computing the shape deformation as the Euclidean distance between the object boundary points and the corresponding points in the model cross section boundary. The shape deformation distances measured in different cross sections are average and the minimum average shape deformation distance is used to identify the model best matching to the object of unknown classification.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Image position and motion detecting method

Abstract: PURPOSE:To detect the position and motion of an object with high accuary even when plural objects exist or a background is moved by extracting a partial area provided with desired color feature quantity in block unit based on relation between the chrominance signal of a picture element and a degree of assignment for the set of desired color feature quantity of the object. CONSTITUTION:The degree of assignment for the set of the desired color feature quantity is found from the chrominance signal of each picture element in image data, and the partial area provided with the desired color feature quantity is extracted using the degree of assignment, and the position of the area is found in rectangular block unit. For example, both the sum(total frequency) S0 of the degree of assignment of a block b0 and the sum S1 of a block b1 show values over a reference value, and the blocks b0 and b1 comprise a continuous block group relating to the area 1. The center position of the area 1 can be found by the centroid of the blocks b0, b1 and the sums S0, S1. In such a way, the position, of the partial area of the image data can be detected by extracting the pararea 1 and finding the center of it.

Tracking area setting device, motion vector detecting circuit and object tracking device using the same

Abstract: PURPOSE: To clearly and automatically distinguish an object from the other area CONSTITUTION: A detection block to be used is set in a mode 1 or 2 Based on the representative point data of a detection block latched in a preceding field and the pixel data of a detection block latched in a present field, a minimum correlation value is calculated by a correlator and based on that value, a partial motion vector is calculated and defined as a reference vector In the next field, based on the partial motion vector and the reference vector for each detection block, a correlation value is provided by the correlator When that correlation value is smaller than a threshold value, it is defined as a detection block consisting of a tracking area by a detection block deciding circuit The centroid of the tracking area is calculated by a centroid detection circuit and based on that centroid, a motion vector is outputted from a motion vector generating circuit The reference vector is updated for each field Based on the motion vector, a universal head control circuit 16 controls a driving circuit 18 and a universal head 20 is driven by the driving circuit 18 so that a camera 22 can track the object COPYRIGHT: (C)1996,JPO

Using the centroid operator for faster multiresolution image compression and pattern recognition

Abstract: It is possible to build a multiresolution image encoding technique using the fractal transform. Fractalencoding methods rely on measuring least mean square difference between image blocks at two different spatialscales for most efficient block matching. If we instead, compute the centroid of the blocks to be matched beforematching occurs and preserve this centroid information in the encoding process we both speed up the encoding pro-cess and have information that will allow us to interpret the shape of objects in the encoded scenes. This ability maydramatically increase the speed at which pattern identification in large volumes of image data may be performedsince less data would have to be processed for searches over large number of images.Keywords: centroid, multiresolution, image compression, pattern recognition, fractal 1. INTRODUCTION Jacquin5 began the process of the fractal transform by dividing an object into two sets of blocks. Therange blocks. as he describes them, are square blocks that are some integral power of two in size and the domainblocks are square blocks that are the next larger power of 2 in dimension. Typically range blocks are 4*4 or 8*8 pix-els in size. Matching is achieved by decimating the domain blocks by a factor of two and using the least mean squaresdifference between a given range block with all possible domain blocks with 8 possible flips and rotations applied tothem (see Fig. 1). In this way, matching occurs between two different scales of detail in the image. This algorithm isconcisely described by Bogdan1 with

Method for estimating position and attitude of three-dimensional object

Abstract: PURPOSE:To stably and efficiently determine the positional and attitude of a three-dimensional object based upon range data even when three-dimensional attitude variation exists or an aspect ratio is close to '1'. CONSTITUTION:A multi-direction histogram string for a model, centroid on a vertical plane in respective directions and an angular direction centroid map are previously calculated. While observing an object, a histogram, centroid on a vertical face in an object direction and an angular direction centroid map around the centroid are calculated. Three degrees of freedom for the object direction and position out of six degrees of freedom are determined based upon correlation between the histogram and the multi-direction histogram data sequence, two degrees of freedom for a position on the vertical plane in the object direction are determined based upon the difference of centroid between the model and the object and one degree of freedom for a rotational angle to the object direction is determined based upon the correlation of angular direction centroid maps between the model and the object to improve the efficiency of calculation. This estimating method can be applied also to an object whose aspect ratio is close to '1' based upon correlation between a histogram and an angular direction feature amount or to a non-projected object or an object having three-dimensional attitude variation based upon correlation between a histogram and a multi-direction histogram string.

A linear current to position converter for "CDD" functional design

Abstract: The paper presents a method for a current to position converter that provides a linear dependence between the gap of two currents and centroid position of a charge domain on a given contour. A numerical method is used to model the potential distribution on the edge of a resistive layer where a current is injected. The optimal shape of the layer and the position of the injection points are determined.

Method and device for extracting target object in image

Abstract: PURPOSE:To separately extract each object in the case plural objects are picked up in a superposed state CONSTITUTION:An initial area candidate extracting part 2 extracts a candidate of an initial area for extracting a target object, and an initial area extracting part 3 obtains the initial area from the initial area candidate An object position calculating area selecting part 4 selects a calculation area in the corresponding object position from in the initial area candidate A centroid calculating part 5 derives the centroid of an area of a world coordinate system corresponding to the object position calculation area An object position calculation area examining part 6 examines whether the area of the world coordinate system corresponding to the object position calculation area is contained within a prescribed range from the centroid obtained by the centroid calculating part 5 or not In the case it is decided that the area of the world coordinate system corresponding to the object position calculation area is contained within the prescribed range from the centroid, by the object position calculation area examining part 6, a target object extracting part 7 sets the centroid as a position of a target object, and extracts a picture element corresponding to the area existing within the prescribed range in the world coordinate system, as the target object from a position of the target object

Change area integral method for monitor picture

Abstract: PURPOSE:To attain extract processing of an accurate characteristic quantity by smoothing a histogram projecting difference binarized picture with a centroid coordinate of a maximum area as a center and integrating the maximum area and areas in the vicinity of the maximum area into one area. CONSTITUTION:A difference processing section 4 obtains an absolute value of a difference of luminance between an input picture and a background picture and provides an output of a changes as a difference picture. A binarizing processing section 5 obtains the difference binarized picture. A labelling processing section 6 gives a label to a change area. A maximum area detection section 7 detects an area whose area is maximum. A centroid coordinate calculation section 8 obtains the centroid coordinate of the maximum area. A projection histogram smoothing section 10 obtains a maximum value and a minimum value of X and Y axes while smoothing a projected histogram of the difference binarized picture in the X and Y axis directions. A re-labelling processing section 11 gives a same label to a label of a change area in a rectangular area decided by them. Thus, an integral processing section 12 integrates an area in the rectangular area into one.