International Business Machines Corporation

Charles A Micchelli

Marco Martens

Peder Andreas Olsen

Sankar Basu

Joan LaVerne Mitchell

Chai Wah Wu

Charles Philippe Tresser

Gordon Wesley Braudaway

Timothy James Trenary

Frank P Giordano

Pages of books are copied without distortion due to curvature of the page near the book binding or the distortion in a copied page is corrected using the spacing of equidistant bars on tape strips applied to the top and bottom edges of a page before copying. The tape is preferably transparent and rather narrow and easily attached to a page to be copied. The first step in the distortion correction procedure is to locate the bars at the top and bottom of the page. The distortion of the spacing between the imaged bars is computed based on the known distance between the equidistant bars. The computed distortion of the spacing is then input to a distortion correction algorithm. The output of the distortion correction algorithm generates a corrected image. This image may also optionally delete the bars so that they are not printed in the copy. The corrected image is then copied.

Method and apparatus to correct distortion of document copies

A parametric family of multivariate density functions formed by mixture models from univariate functions of the type exp(&minus;&verbar;x&verbar; ) for modeling acoustic feature vectores are used in automatic recognition of speech. The parameter &bgr; is used to measure the non-Gaussian nature of the data. &bgr; is estimated from the input data using a maximum likelihood criterion. There is a balance between &bgr; and the number of data points that must be satisfied for efficient estimation.

Impulsivity estimates of mixtures of the power exponential distrubutions in speech modeling

A one-dimensional algorithm for perform the merging of complementary portions from two independent overlapped images on the same 8&times;8 grid without the computational expense of conversion to and from the real domain is extended to a two-dimensional procedure. The merging process is performed exclusively in the frequency domain. The extension to two dimensions is done by shifting and/or merging rows of side-by-side two-dimensional transformed data blocks using the one-dimensional algorithm. Then a vertical shifting and/or merging can be performed on the horizontally shifted and/or merged blocks again using the one-dimensional algorithm with an independent shift/merge parameter.

Shift and/or merge of transformed data along two axes

A method of transform domain processing of transformed data is applied to image processing in various applications. A one-dimensional algorithm allows for the selection of an arbitrary contiguous eight component DCT sub-block from two adjacent DCT blocks that foregoes the expense of image manipulation in the real domain. Due to the fact that non-zero DCT coefficients are generally sparse, this algorithm lends itself nicely to the development of special cases which are faster than methods in use today. The de-quantization and re-quantization can be combined at execution time with the constants.

Shift and/or merge of transformed data along one axis

Improvements in speech recognition systems are achieved by considering projections of the high dimensional data on lower dimensional subspaces, subsequently by estimating the univariate probability densities via known univariate techniques, and then by reconstructing the density in the original higher dimensional space from the collection of univariate densities so obtained. The reconstructed density is by no means unique unless further restrictions on the estimated density are imposed. The variety of choices of candidate univariate densities as well as the choices of subspaces on which to project the data including their number further add to this non-uniqueness. Probability density functions are then considered that maximize certain optimality criterion as a solution to this problem. Specifically, those probability density function's that either maximize the entropy functional, or alternatively, the likelihood associated with the data are considered.

Maximum entropy and maximum likelihood criteria for feature selection from multivariate data

A nonparametric family of density functions formed by histogram estimators for modeling acoustic vectors are used in automatic recognition of speech. A Gaussian kernel is set forth in the density estimator. When the densities are found for all the basic sounds in a training stage, an acoustic vector is assigned to a phoneme label corresponding to the highest likelihood for the basis of the decoding of acoustic vectors into text.

Penalized maximum likelihood estimation methods, the baum welch algorithm and diagonal balancing of symmetric matrices for the training of acoustic models in speech recognition

A statistical modeling paradigm for automatic machine recognition of speech uses mixtures of nongaussion statistical probability densities which provides improved recognition accuracy. Speech is modeled by building probability densities from functions of the form exp(&minus;t,) for t&gE;0 and &agr;>0. Mixture components are constructed from different univariate functions. The mixture model is used in a maximum likelihood model of speech data.

Growing community of inventors

Mohegan Lake, NY, United States of America

Charles A Micchelli