Image analysis – Applications – Reading bank checks
Reexamination Certificate
1995-02-15
2001-12-04
Chen, Wenpeng (Department: 2624)
Image analysis
Applications
Reading bank checks
C382S207000, C382S320000, C235S449000
Reexamination Certificate
active
06327378
ABSTRACT:
FIELD OF THE INVENTION
The present invention pertains to a recognition method for printed characters on documents and the like wherein an electrical signal from a character reading device is analyzed on an amplitude versus time basis by sampling the signal waveform at selected intervals and comparing sets of values representing ratios of amplitude direction or slope, or comparing sets of values of differences of amplitude ratios with predetermined recognition templates for characters expected to be read.
BACKGROUND
Several methods and systems have been developed for automatically reading printed characters on documents for document processing and record purposes. For example, in processing bank checks through the banking system, standards for check identification characters have been well established and systems and methods have been developed for optically or magnetically scanning the printed characters on the document for identifying the characters and recording the identified information for various purposes. The banking system, for example, has developed a standard for printed characters known as E13B wherein the characters are printed with ink which may be magnetized and then read by a magnetic flux sensor. Alternatively, the characters may be printed with sufficient contrast with respect to the document background to enable optical scanning.
Magnetically scannable characters, as well as optically scannable characters, often have imperfections which cause the reading devices to produce signal defects. Check documents, for example, have historically been printed by offset, lithographic or ribbon film techniques, each of which produces occasional defects in the printed character which interfere with the recognition process. These character defects show up in the analog waveform signal which is generated by the character reading device. Defects such as the overall waveform amplitude, shift of signal peaks and extraneous secondary signal peaks are sometimes generated. These defects interfere with the character recognition method wherein the scanned waveform is compared with a set of templates which represent all of the possible characters being read on the document.
Although certain adjustments have been made to character template sets in automatic character recognition systems to accommodate signal defects due to various printing techniques, additional printing methods and equipment or system peculiarities have created new signal defect problems in prior art character reading systems and methods. For example, in recent years the use of electrostatic printing techniques using so-called toner print mediums have caused character recognition signal defects including waveform peaks that have greater distortion in relation to other peaks in the character in both magnitude and position, the tendency of the waveform to gradually fall to the signal base line as compared to a rapid signal fall off even though the character has passed the scanning head, and weaker initial waveform peaks which are crucial to the determination of the initiation of reading a character.
Various techniques have been developed in the art of optical and magnetic character recognition. For example, techniques have been developed for reading the amplitude versus time analog signal of a character reading device wherein sampling of the signal polarity versus time at selected intervals across the width of the character forms a set of polarity values which may be compared to a template. Systems and methods have been developed where a time based amplitude signal is sampled at several selected time intervals, the total of which corresponds to the width of the character, the analog signal is converted to a set of binary characters and then compared to binary character sets corresponding to the characters expected to be read.
More recently, methods have been developed wherein the “volume” of ink within sections of the character has been measured and compared to predetermined ink volume character sets for character identification. In this regard, the amplitude of an electrical analog signal generated by the reading device represents the “volume” of ink of the height of the character under the scanning aperture. Typically, the input signal from the magnetic or optical reading device or “head” is amplified and converted from an analog waveform to a series of digital words at periodic intervals across the character. The series of digital words, representing the full character width, are summed to create an average and each individual word is divided by the average to generate a new digital word that represents the ratio of the individual digital words to the average. These ratios are then sequentially matched against a set of character templates for recognition purposes. Typically, the templates consist of a series of ratio ranges for each of the samples in the character series. Each ratio of the series is compared to the template ranges for each character. If all of the ratios are within the template range for a given character, that character is then recognized.
Typically, the template ranges are designed so that all characters have at least one template point that is mutually exclusive from all the other characters to prevent two characters from being simultaneously recognized as one or the other. However, the above described technique is based on measuring signal amplitude at each measurement point and then compared to an average signal amplitude across the width of the character. This technique is thus subject to certain reading errors when reading characters printed according to methods which tend to generate signals of greater or distorted amplitudes across the character width.
The following references have been identified as representing the state of the art in magnetic or optical character recognition methods, including those discussed hereinabove.
Patent No.
Inventor
Issue Date
3,114,131
R. J. Furr et al.
12/10/63
3,264,609
K. Scheidhauer
08/02/66
3,391,387
I. Flores
07/02/68
3,439,337
C. B. Trimble
04/15/69
3,528,058
M. F. Bond
09/08/70
3,571,793
R. H. Britt
03/23/71
3,629,829
R. Ordower
12/12/71
3,638,238
R. E. Milford et al.
O1/25/72
3,641,495
K. Kiji
02/08/72
3,818,446
D. A. Benson
06/18/74
4,032,887
D. C. Roberts
06/28/77
4,245,211
C. T. Kao
01/13/81
4,315,246
R. E. Milford
02/09/82
4,547,899
R. B. Nally et al.
10/15/85
4,827,531
R. E. Milford
05/02/89
5,026,974
G. R. Franklin et al.
06/25/91
5,257,319
Canu et al.
10/26/93
Contrary to the systems and methods represented by the above-cited references, the present invention is directed to a character recognition system and method which overcomes some of the problems associated with variations in magnetic ink consistency, interaction of the magnetized or contrasted character with the reading device and its associated circuitry and certain printing techniques including those which use electrostatic toner type ink materials which cause signal amplitude distortions.
SUMMARY OF THE INVENTION
The present invention provides an improved method for recognizing printed characters on a document or the like by treating an amplitude versus time representation of a signal generated by a character reading device in a manner which overcomes certain signal defects caused by printing methods and certain anomalies of the reading device and associated circuitry.
In accordance with one aspect of the present invention, a character reading method is provided wherein an amplitude versus time analog signal or waveform representing each character is measured at selected time intervals across the width of the character and the signal values at the sampling intervals are subjected to unique analysis and compared to a set of templates representing all of the characters expected to be read by the reading system.
In one embodiment of the invention, amplitude waveform is measured or sampled at selected time intervals, a set of amplitude magnitude values is averaged, each sample of the set is divided by the average for the set to provide a set of amplitude ratio values. The magnitude of the
BancTec Inc.
Chen Wenpeng
Glaser Kenneth R.
Martin Michael E.
LandOfFree
Character recognition method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Character recognition method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Character recognition method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2564611