An adaptive approach to offline handwritten word recognition

Abstract

AbstractÐAn adaptive handwritten word recognition method is presented. The key ideas of adaptation are 1) to actively and successively select a subset of features for each word image which provides the minimum required classification accuracy to get a valid answer and 2) to derive a consistent decision metric which works in a multiresolution feature space and considers the interrelationships of a lexicon at the same time. A recursive architecture based on interaction between flexible character classification and deductive decision making is developed. The recognition process starts from the initial coarse level using a minimum number of features, then increases the discrimination power by adding other features adaptively and recursively until the result is accepted by the decision maker. For the computational aspect of a feasible solution, a unified decision metric, recognition confidence, is derived from two measurements: pattern confidence, evaluation of absolute confidence using shape features, and lexical confidence, evaluation of the relative string dissimilarity in the lexicon. Practical implementation and experimental results in reading the handwritten words of the address components of US mail pieces are provided. Up to a 4 percent improvement in recognition performance is achieved compared to a nonadaptive method. The experimental result shows that the proposed method has advantages in producing valid answers using the same number of features as conventional methods.Index TermsÐPattern recognition, handwritten word recognition, adaptive word recognition.

Marathi Numeral Identification System in Devanagari Script Using 1D Discrete Cosine Transform

“…Fuzzy logic also can be used for recognition purpose [19]. Park J worked on whole word recognition without segmenting it [20]. Document Text retrieval without using optical character recognition also can be performed [21] along with the minimization in the error rate [22].…”

Multistage Recognition Approach for Offline Handwritten Marathi Script Recognition

“…Jenowa Park Proposed offline handwritten word recognition(HWR) [9].The key ideas are actively and successfully select a subset of features for each word image which provides the minimum required classification accuracy to get a valid answer and to derive a consistent decision metric which works in multi resolution feature space and considers the interrelationship of lexicon at the same time .A recursive architecture based on interaction between flexible character classification and deductive decision making is developed. The recognition process starts from the initial coarse level using a minimum number of features, then increase the discrimination power by adding other features adaptively and recursively until the result is adapted by the decision maker .For the computational aspect of a feasible solution ,unified decision metric ,recognition confidence ,is derived from two measurements: pattern confidence, evalution of absolute confidence using shape features and lexical confidence evaluation of the relative string dissimilarity in the lexicon He has implemented the same for the US MAIL Address component.…”

“…Median Filtering has been employed, to remove irregularities such as ''speckle noise", "salt and pepper noise" in the digital image introduced in the scanning process .Each pixel has been replaced by the median of the neighbouring pixels. [9] f (x, y) = median {g(s, t)}.…”

“…Let D be the distance from C to the edge of the neighborhood. The neighborhood is then all (X,Y) Such that C-D < X < C+D And C-D < Y < C+D (9) This defines a square neighborhood about c. Since the width of the neighborhood decreases over the training iterations, the value of d decreases. Initially d is set at a chosen value denoted by d 0 may be chosen at a half or a third of the width of the competitive layer of the processing units [2] [4].…”

Performance Evalution of Multistage Offline Marathi Script Recognition System

“…Proposed offline handwritten word recognition(HWR) [9].The key ideas are actively and successfully select a subset of features for each word image which provides the minimum required classification accuracy to get a valid answer and to derive a consistent decision metric which works in multi resolution feature space and considers the interrelationship of lexicon at the same time .A recursive architecture based on interaction between flexible character classification and deductive decision making is developed. The recognition process starts from the initial coarse level using a minimum number of features, then increase the discrimination power by adding other features adaptively and recursively until the result is adapted by the decision maker .For the computational aspect of a feasible solution ,unified decision metric ,recognition confidence ,is derived from two measurements: pattern confidence, evaluations of absolute confidence using shape features and lexical confidence evaluation of the relative string dissimilarity in the lexicon He has implemented the same for the US MAIL Address component.…”