GB2419724A - Obtaining information about the honest intent of handwriting - Google Patents

Abstract

An electronic representation of handwriting is scanned to determine the presence of one or more predetermined characteristics common to handwriting of more than one person and indicative of the handwriting having been written with dishonest intent. The intent is classified as potentially dishonest if the number and/or type of characteristics(s) detected satisfies predetermined conditions.

Description

METHOD AND APPAPATUS FOR OBTAINING INFORMATION

ABOUT THE HONEST INTENT OF HANDWRITING

The invention relates to methods and apparatus for obtaining information about the honest intent with which a sample of handwriting was written.

The invention is particularly concerned with the analysis of handwritten information in application forms and the like including, for example credit card applications, social service claim forms, insurance claims and employment applications where the author of the handwriting can be easily verified, for example because the writing is provided in the presence of another person, but the honesty of the content of the writing cannot be verified.

Computer software algorithms are well known which enable a sample of handwriting to be analysed in order to determine its authenticity. This is done by comparing the new sample of writing to a known one (for example a signature) stored in a database. The two samples are compared for different characteristics, for example pressure strokes, speed of writing, shape of letters etc. While the prior art can identify samples written by person "A" pretending to be person "B", they cannot identify samples that have been written by person A but with dishonest intent. That is for example there is no attempt to disguise the handwriting or identity of the writer (although sometimes this may apply) but to disguise the dishonest intent of the writer.

In accordance with the first aspect of the present invention, a method of obtaining information about the honest intent with which a sample of handwriting was written, the method comprises generating an electronic representation of the handwriting; and scanning the electronic representation of the handwriting to determine the presence of one or more predetermined characteristics common to handwriting of more than one person and indicative of the handwriting having been written with dishonest intent; and classifying the intent as potentially dishonest if the number and/or type of said characteristic(s) detected satisfies predetermined conditions.

In accordance with a second aspect of the present invention apparatus for obtaining information about the honest intent with which a sample of handwriting was written comprises an imaging system for obtaining an electronic representation of the handwriting; and processing means for scanning the electronic representation of the handwriting to determine the presence of predetermined characteristics common to handwriting of more than one person and indicative of the handwriting having been written with dishonest intent, and for classifying the intent as potentially dishonest if the number and/or type of said characteristic(s) detected satisfies predetermined conditions.

We have realised that it is possible to utilize certain techniques known from the field of graphology in a partially or fully automated manner to obtain information about the honest intent of handwriting. It should be understood, however, that the invention is not designed to provide a fool proof indication of whether a sample of writing was written with honest intent or not, but rather to identify those samples that require further investigation to ascertain whether they are indeed honest.

The present invention seeks to identify those samples of writing that are potentially dishonest (even where they are written by the correct person) for further investigation. Dishonesty or dishonest intent in this context refers to the situation when someone, who is not hiding his indentity, writes a statement and they know it to be false or dishonest. This may for example include claiming too much on an insurance claim form, or putting incorrect information on a credit card application.

The science of graphology has shown that dishonesty such as this - whether it is a one-off or persistent event - is detectable in handwriting by the presence of certain "attributes" or characteristics. These attributes are associated with particular personality traits, for example deception, not telling the truth, concealing information or having criminal tendencies.

While the general idea may be applied to a range of situations, including the analysis of applications for refugee status, writing used to gain entry to a building or access to a service such as an article dispenser, for example cash dispenser, this description will concentrate on dishonesty in regard to credit card, ID card, passport, driving licence, house sales/purchases, intelligence checks, employment checks/reference, banking and insurance forms. This type of dishonesty is subtler (i.e. a form of fraudulent dishonesty) and as a consequence the attributes to look for are more limited. (The attributes in other cases will be different) Other applications of the invention include the analysis of information provided in handwriting at airports, seaports, and customs and excise.

In order to analyse the handwriting, it is necessary to determine the presence of certain predetermined characteristics or attributes and we have recognised that there is a relatively small number of characteristics of handwriting which are indicative of dishonest intent. It is relatively straightforward to determine the presence of this limited number of characteristics and thus a rapid analysis of the handwriting can be performed.

Although the predetermined conditions may require that just one of the said characteristics is determined, typically, the predetermined conditions are satisfied if the number of detected characteristics is greater than a non-unity threshold. We have found that the presence of at least 4, preferably 5, most preferably 7 characteristics is usually enough to indicate potential dishonesty.

Of course, the conditions can be set depending upon the degree of certainty required with for example more important situations requiring a high degree of certainty and lower level situations requiring a lower threshold.

The electronic representation can be generated by scanning previously generated handwriting or by scanning handwriting as it is being written which increases the chances of detecting dishonest intent since this avoids the problem of someone presenting previously written, but fraudulently obtained, genuine handwriting for scanning.

The scanning process can be carried out using a variety of different conventional equipment such as optical, flatbed scanners, and the like.

Some examples of methods and apparatus according to the present invention will now be described with reference to the accompanying drawings, in which:- Figure 1 is a block diagram of apparatus for performing the method; Figure 2 is a flow diagram illustrating an example of the method; Figure 3 is a flow diagram illustrating another example of the method; Figure 4 is a more detailed flow diagram illustrating an example of the method; and, Figure 5 is a flow diagram of the logical process to accept or reject a document.

The apparatus shown in Figure 1 comprises a flat bed scanner 1 coupled to a microprocessor 2. The microprocessor 2 is coupled with a ROM 3 which stores data, typically a two-dimensional bitmap of each feature characteristic of dishonest handwriting, and to a RAM 4 for storing a digital representation of the writing. The microprocessor 2 is also connected to an output device 5 such as a monitor screen.

In use, a form which has been completed in writing is placed on the scanner 1. The scanner 1 generates a two- dimensional, digital representation of the writing (step 10, Figure 2) and this is supplied to the microprocessor 2 which normalizes the writing so that it takes up a standard size and then the normalized representation is stored in the RAJVI 4. (Step 11) The microprocessor 2 then extracts each bitmap from the store 3 corresponding to each of the dishonesty indicators and compares this using standard pattern comparison algorithms with the digital representation of the signature stored in the RAIS4 4 (step 12) Each time a match is found, an appropriate count is incremented, there being a separate count for each characteristic.

Following the comparison stage, the microprocessor 2 sums the counts and where the sum is four or more for a particular characteristic, it is judged to be present. The total number of characteristics present is then determined and if the total is greater than a predetermined threshold such as 7 (step 13) the microprocessor 2 generates a reject message on the monitor 5 (step 14) and/or issues an audible tone or other output indicating the probability that the intent is potentially dishonest. The form is then passed to an investigator for manual checking. The application and/or content may then be considered for further investigation.

Otherwise, the intent is taken to be honest and a "valid" message may be displayed on the monitor 5 (step 15) Following display of the "valid" message, an operator can then validate the form.

In other applications the writing could, if valid, authorise access to a location or service. In addition, the "valid" message, in electronic form, can be transmitted to an adjacent or remote device for storage or further processing.

In Figure 1, the scanner 1 has been shown adjacent to the microprocessor 2. In other applications, this link could take place via a remote communications medium such as a telephone network, Internet or the like. This is particularly suitable for electronic retailing and the like.

In other applications, the scanner 1 could be replaced by a digitizing tablet. The user then completes the form while it rests on the tablet allowing the writing to be monitored as it is written. This also allows pressure characteristics to be detected.

A second form of apparatus which could be used to implement this invention is described in detail in US-A- 6,445,820 incorporated herein by reference. Other forms are described in www.garthrnichaels.com and/or www. handwritinganalysis. corn.

To identify the attributes or characteristics most commonly associated with dishonest intent a range of application forms that are known to have been written with dishonest intent are analysed. As a result, a list of key attributes and their importance (as measured by frequency in known samples) can be drawn up. Attributes may be "primary dominants" i.e. those that are seen when looking at the paper as a whole, for example the size of the writing, slant, pressure and word spacing; "secondary dominants" such as writing with joined and detached letters, simplified script or variable writing; and "miscellaneous features", for example starting strokes, end strokes, loops in letters like "d", "k" "1" and "r" and the appearance of "o"s and "a"s.

Attributes may be present singly or in combination with each other. While some attributes are overwhelming indicators of dishonesty on their own, in certain instances they may not be. For example writing sloping to the right indicates an individual that is forceful in their opinion and will get their way - not a direct indication of dishonesty - but with unusually small writing, it can signify an individual who is very tense and will do whatever necessary to get their way. This could for example include being dishonest in order to get a new bank loan or credit card.

Attributes typically must also occur a number of times in a sample of writing before they can be considered S genuine indications of a person's character. For example, typically upside down ovals are indications of underhandedness, but one or two instances of these in a sample of writing could be a coincidence, while four or more indicate a general tendency.

A list of suitable attributes is provided in the Appendix.

The preferred method used by the processor 2 to analyse a document is shown in Figure 3 and follows the Hilliger method currently used by graphologists.

Layout involves looking at the overall symmetry of the script, and an easy way to do this is by dividing a sample of writing into different zones upper, middle and lower (as understood by someone skilled in the art) . Each zone is measured individually and a general assessment of the script is thereby obtained.

In the case of a page of handwriting, overall symmetry implies the general arrangement of the text on the paper.

For example are there sufficient and not exaggerated margins at the top, bottom, right and left edges of the paper and is the general effect of the page balanced and harmonious? To gauge overall symmetry the following features may be assessed: - Words and letters. The horizontal spacing or intervals between words is assessed i.e. are the blank spaces between words of approximately equal length? Is the spacing between words adequate for clarity? Is the spacing between the letters too close (horizontal mingling), adequate, or too wide (secondary/artificial broadness)? Lines. Is the overall vertical spacing between the lines balanced in relation to the absolute size of the handwriting and the size of the paper? Speed is assessed by evaluating a number of dominants: 1. Degree of connectedness of letters within words.

"Clever linking" increases the speed of the act of writing and so is "quicker", very disconnected writing = slow.

2. Forms of connection i.e. "Copybook", "Angularity" and "Arcades" are slow, while "Garlands", "Wavy-lines" and "Threads" are quick.

3. Degree of broadness - small letters "n" and "u" are used as guidelines to establish broadness. Covering strokes (where the stroke retraces one already made) = extreme narrowness. Pseudo-broadness = wide connecting strokes between narrow lines. Horizontal mingling i.e. letters touching each other in the middle zone = narrowness.

4. Pressure. Light = quick (less resistance) and heavy = slow. Distinct pressure (heavy down and light up) enhances speed. Indistinct pressure lacks an even flow so = slower.

5. Rhythm and regularity. Intelligently regular and rhythmic writing is considered quick, whereas monotonous copybook regularity is slow. Irregular writing is quick.

6. Diactitics i.e. "i" dots, "t" crosses, accents etc. Slow = "1" dots and "t" crosses placed left of the down stroke, or placed accurately in a leftward-slanting writing, triangular or knotted "t" crosses, interrupted writing in the middle of a word to add diacritics. Quick = rightward diacritics, "clever-linking" ( , "i" dots placed high and to the right or made into dashes etc. 7. Slant. Upright and/or left slant = slow, right slant = quick. Loops in the lower zone "pulled back" so as to become leftward slanting at that point, or small letters in the middle zone and that slant to the left in a predominantly right slanting script are indications of slowness.

8. Simplification and/or enhancement. Elaboration, adornment, embellishment, long starting strokes, circular "1" dots, complicated capitals, enrolments, knots, spirals etc. indicate slowness. Conversely, reduction of the copybook shapes to their basic essential structure and neglected letterforms = quick.

9. Left-hand margin. Narrowing of the left-hand margin down that page = slow, widening = quick.

10. End strokes. Extended or right curve end strokes = quick, non-existent, clipped or downwards end strokes = slow.

11. Direction of baselines. Rising baselines = quick, falling = slow.

12. Distance between baselines in comparison to the absolute size of the writing, Widely separated lines = slow. Closely spaced, crowded or mingles lines = quick.

13. Currency or fluency. Strokes executed smoothly, spontaneously, securely, without any tremors, wavers or wobbles are quick. Narrow margins at the foot of the page or words cramped in at the right-hand margin are also secondary indications of speed. Wavering, bent, twisted or broken strokes, trembling or shaky strokes are all slow.

Frequent pauses, recognisable by meaningless blobs, pen readjustments, divided letters, spasms, spelling mistakes, (amended) touching-up or retraced words, crossings-out etc. all indicate lack of currency and slowness.

14. Pen pressure - available if the writing is monitored as it is written.

When assessing the primary, secondary and miscellaneous attributes the method shown in Figure 4 is used.

In Figure 4 the detector begins by identifying primary dominants 20. If one is found the detector counts 22 the number of times the attribute occurs throughout the sample i.e. the number of instances. If the number is above a threshold value, e.g. 4, the attribute is counted as being present and the detector looks for other "cluster" attributes 24. "Clusters" are groups of attributes that indicate fraudulent dishonesty where one attribute on its own may not.

This process is repeated until the detector has looked for all the primary, secondary and miscellaneous attributes in the database, or until it reaches a threshold value that indicates the document needs to be examined further, whichever is first. The document is then either accepted, if there are no (or few) signs of dishonest intent present, or rejected if there are a large number present and further investigation to establish the veracity of the document is required.

A further example of this method is shown in Figure 5.

Here the detector has been set to the same threshold values as above and only one primary dominant attribute has been found in the same script. In this example no other macro attributes have been found, and the document is accepted if less than five of the cluster searches results in less than four incidences. If all the clusters are present in the document with more than four incidences the document is rejected.

In the examples, the handwriting is stored as a 2D bit map. In other cases, the handwriting could be stored as a grey scale or even colour to provide an indication of the pen pressure applied, the darker the line the more pressure. This, in combination with line width, could replace the need for pressure sensitive pads and support the recognition of the characteristics such as "Overt Heavy Pressure" without the need for live capture of the handwriting.

Claims (16)

CLLIMS

1. A method of obtaining information about the honest intent with which a sample of handwriting was written, the method comprising generating an electronic representation of the handwriting; and scanning the electronic representation of the handwriting to determine the presence of one or more predetermined characteristics common to handwriting of more than one person and indicative of the handwriting having been written with dishonest intent; and classifying the intent as potentially dishonest if the number and/or type of said characteristic(s) detected satisfies predetermined conditions.

2. A method according to claim 1, wherein the predetermined conditions are satisfied if the number of detected characteristics is greater than a threshold.

3. A method according to claim 1 or claim 2, wherein a predetermined characteristic is present if more than three occurrences are detected.

4. A method according to any of claims 1 to 3, wherein the handwriting comprises entries on a form.

5. A method according to any of the preceding claims, wherein the electronic representation is generated by scanning previously generated handwriting.

6. A method according to any of claims 1 to 4, wherein the electronic representation is generated by monitoring handwriting as it is being written.

7. A method according to any of the preceding claims, wherein the generating step is carried out remotely from the other steps in the method.

8. A method according to claim 7, wherein the electronic representation of the handwriting is supplied to a remote location for scanning, via a communication medium such as the Internet.

9. A method according to any of the preceding claims, further comprising generating a visual and/or audible output corresponding to the result of the classification.

10. A method of obtaining information about the honest intent with which a sample of handwriting was written, substantially as hereinbefore described with reference to any of the examples shown in the accompanying drawings.

11. A method of controlling access to a restricted service, the method comprising causing a person to submit a handwritten access request; carrying out a method according to any of the preceding claims on the handwritten access request; and, if the handwriting is potentially determined to be written with dishonest intent, denying the person access to the service.

12. A method according to claim 11, wherein the handwritten access request comprises the person's signature.

13. A method according to claim 11 or claim 12, wherein the restricted service comprises an article dispenser such as a cash dispenser.

14. A method according to any of claims 11 to 13, wherein the service is provided from a location remote from the person.

15. Apparatus for obtaining information about the honest intent with which a sample of handwriting was written, the apparatus comprising an imaging system for obtaining an electronic representation of the handwriting; and processing means for scanning the electronic representation of the handwriting to determine the presence of predetermined characteristics common to handwriting of more than one person and indicative of the handwriting having been written with dishonest intent, and for classifying the intent as potentially dishonest if the number and/or type of said characteristic(s) detected satisfies predetermined conditions.

16. Apparatus according to claim 15, the processing means being adapted to carry out a method according to any of claims 1-14.

Appendix _______________________

ATTRIBUTE MEANING

Double looped ovals Deception Stabs in ovals Duplicity Signature different from text Deception Signature contains wide spaces May not genuine Two faced writing Deception Missing out letters Devious by omission Constant mistakes or omissions Person may be lying Ambiguity between letters and Strong money instinct number forms Resting dots Possible deceipt Self conscious strokes Manipulative Slow controlled writing Dishonesty Twisting of letters Manipulation Segmented printing Criminal tendencies Thread writing Manipulative Overt heavy pressure Manipulative Exaggerations tightness Deep seated anxiety Segmented printing Criminal tendencies Exaggerated or disguised Faking writing