International Business Machines Corporation

Scoredata Corporation

Kottappuram Mohamedali Mohiuddin

Jianchang Mao

Prasanta Kumar Behera

Jorma J Rissanen

Tetsunosuke Fujisaki

Raymond Amand Lorie

Richard G Casey

Vasudev Bhandarkar

A computing system is provided that includes a runtime (RT) controller that produces a corresponding sequence of individual RT prediction results. A FIFO buffer organizes RT confidence score values into RT confidence score value frames. An RT storage circuit stores RT confidence values in an RT value histogram structure that indicates frequency of occurrence of individual RT confidence score values. A training controller produces training prediction results. A training storage circuit stores training confidence values in a training value histogram structure that indicates frequency of occurrence of training confidence score values. A comparison circuit compares RT confidence values with training confidence values in the training value histogram structure. A calibration controller circuit produces a calibration update signal based upon the comparison.

System and method for run-time update of predictive analytics system

Fixed-pitch, fixed-font characters embedded in a noisy gray-scale image of picture elements (pels) within a complex background can be extracted prior to execution of any recognition operations by first deriving a normalized Boolean-coded image from the gray-scale image. Then, a subset of at least three uncontaminated character triples is formed by filtering the Boolean-coded image. Next, an affine transform is approximated from locations in the Boolean-coded image of at least three noncollinear ones of the uncontaminated character triples. Lastly, the locations in a logical matrix array of all possible character triples are estimated according to the affine transform.

Method and means for extracting fixed-pitch characters on noisy images

A multi-stage multi-network character recognition system decomposes the estimation of a posteriori probabilities into coarse-to-fine stages. Classification is then based on the estimated a posteriori probabilities. This classification process is especially suitable for the tasks that involve a large number of categories. The multi-network system is implemented in two stages: a soft pre-classifier and a bank of multiple specialized networks. The pre-classifier performs coarse evaluation of the input character, developing different probabilities that the input character falls into different predefined character groups. The bank of specialized networks, each corresponding to a single group of characters, performs fine evaluation of the input character, where each develops different probabilities that the input character represents each character in that specialized network's respective predefined character group. A network selector is employed to increase the system's efficiency by selectively invoking certain specialized networks selected, using a combination of prior external information and outputs of the pre-classifier. Relative to known single network or one-stage multiple network recognition systems, the invention provides improved recognition, accuracy, confidence measure, speed, and flexibility.

Concurrent two-stage multi-network optical character recognition system

A method of creating a decision tree to enable a character recognition device to recognize characters in a new unknown font includes scanning a document printed in the unknown font to generate an array of pixels, with each pixel having a neighborhood state based on the colors of the surrounding pixels. A plurality of clusters of pixels, each representing a printed character, are identified. A pixel in one cluster is selected, and its neighborhood state determined. The neighborhood state is used to address a memory that has stored in it a probability table providing the probability, in a second font different from the new font being taught, that a pixel is black as a function of its neighborhood state. The stored probability of black associated with the neighborhood state of the select pixel is read from the memory and assigned as the probability of the selected pixel. The decision tree for the new font is generated using that assigned probability of black.

Method and apparatus for establishing pixel color probabilities for use

Method and apparatus which cyclically generate a compressed, arithmetically-coded binary stream in response to binary occurrence counts of symbols in an uncoded string. The symbols in the uncoded string are drawn from a multi-character alphabet which is not necessarily a binary one. Coding operations and hardware are simplified by deriving from the binary occurrence counts an estimate of the probability of each unencoded symbol at its precise lexical location. The probability estimation eliminates any requirement for division or multiplication by employing magnitude-shifting of the binary occurrence counts. The encoded stream is augmented by the estimated symbol probability at the same time that an internal variable is updated with an estimate of the portion of a probability interval remaining after coding the current symbol, the interval estimate being obtained from the left-shifted occurrence counts. Decoding is the dual of encoding. The unencoded symbol stream is extracted, symbol-by-symbol, by subtracting the estimated symbol probability that comes closest to, but does not exceed the magnitude of the compressed stream, re-estimating the symbol probabilities based upon the decoding, and testing the difference of the subtraction against the re-estimated probability.

Growing community of inventors

San Jose, CA, United States of America

Kottappuram Mohamedali Mohiuddin