-The paper reveals the superficial matching between script features as understood by forensic experts and computer scientists and advocates the development of computational instruments tailored to fit the features traditionally used by the forensic community. In particular, and including other areas of graphonomics and the general public, there exists a demand for software for the analysis of intuitive features, think “slant” or “roundness,” as opposed to analytical features, like “Fourier transform” or “entropy.” Rex, a software with such a capability, is introduced and used to explore the potentialities of this approach for script forensics. An investigation of properties of the script contour orientation, the feature used by Rex, is also presented.
Index Terms—script features, contour orientation, computational graphonomics, handwriting forensics,
Semiotics — That much forensic handwriting expertise is
subjective and would profit from mathematics and computing
in its quest for objectivity and replicability is publicly admitted
[
Cognition — An interesting viewpoint on the debate over intuitive and analytic features is to consider mathematics as an evolutionary outcrop of the neural computing capacities of the brain. Intuition is evolutionary unconscious learning by interaction with the environment to which conscious analysis supplements when novelties arise. Thus the two can be envisioned as a continuum, mathematics progressively becoming intuitive.
Sociology — To think that the divergence of the two feature types is a function of mathematical educational level is overlooking a fundamental distinction. Writer identification and verification are main mobiles of computational handwriting forensics, and because here only results count, it can use any method without even the need of thorough understanding insofar as it is better. This evolutionary mindset of a goal-focused black box approach is faced by the knowledge-oriented crystal ball attitude seen in the traditional graphonomical research, which adds to the control tasks mentioned above a considerable interest in the handwriting ecosystem, i.e. the structures and dynamics of handwriting features across populations and the underlying factors: material, cognitive, biomechanical, sociocultural.
Linguistics — The issues with the term “feature” extend to
a further worldview cloaking inconspicuously its users. The
proposition “This font is Roman” is considered in philosophy
either as an expression on a property owned by the font
(objectivism) or attributed to the font by an observer (subjectivism)
[
Implications — Computer scientists have to consider in common intelligence with forensic experts three issues worth mentioning because they bear an influence on how the software presented later in the paper is to be used. The issues are the desired precision of the analysis, the definition of the features and the affordability to analyze them in the current state of the art. I will illustrate this through two visual examples.
Precision — Fig. 1 presents three bitmap circles of various
sizes for which the orientation along their contour is measured
(details in section iii). Being circles, we would expect that all
orientations be equally well represented, but due to the discrete
nature of the underlying raster in which the shapes live the
distribution is biased towards the orthogonal direction — the
distribution will peak at 0 and 90 degrees ([
Definition — I discuss now the slant of three Roman script characters as perceived by a human and raise the question of how this simple feature should be defined. In the case of I the slant is vertical and corresponds to the shape’s axis of equilibrium through its center of gravity — here the slant is a physical property of the object. For an O there is no way to tell how the character is oriented would the baseline be unknown — slant is here a property of the object relative to the surrounding. The slant of y can be considered as upright only if we are able to identify the shape as character “y” and be aware of the convention that this lower case letter has to be considered vertical despite its physical right-leaning — this is a case of semantic slant. A deeper examination might reveal even more criteria. In conclusion, a slant analysis algorithm implementing human expert behavior appears to be more challenging than suspected, given first the very difficulty to define the feature, and secondly due to the mix of perceptual and cultural considerations to model.
Afordability — The last sentence leads to the issue of affordability: do we have the technological means to perform comprehensive slant analysis since we need to recognize unconstrained handwritten characters? This task not being presently solved, a positive answer can be given only if we are happy with a certain degree of imprecision, its exact amount having to be determined. Some of the fine computational forensic expertise that we would wish to attain is thus yet out of reach.
Rationale — Written documents in databases can be retrieved
by appearance by one of the following methods: visual (using
a reference document), semantic (describing script features),
haptic (by drawing) and exogenous (from document ecosystem
metadata). Semantic retrieval is convenient because it is intuitive
(it takes place via a graphical and natural-language interface),
free of any preexisting model (not always available) and can
describe aspects of a script (contrary to the holistic approach of
visual retrieval). The software that grew out of these
considerations, called Rex, suits the demand for tools supporting forensic
specific features as described above (Fig. 2) [
Technicalities — The software measures the local orientation
along the writing contour, a popular computational
graphonomics feature [
Applications — The Swiss reference is not fortuitous, since the handwriting documents presently used by Rex originate in that country (IAM Handwriting Database 3.0 [20]). This shows again the surprising versatility of the tool in that it is not only a document browser, but also a teaching tool about handwriting. In addition to learning about individual documents, Rex provides an insight in the make-up of a population of writers — that of the canton of Bern from where most of the dataset writers hail (Fig. 4). The question that immediately springs to mind —“Do writers from other parts of the world have similar characteristics?” — is typical of the richness of research and pedagogical possibilities opened by such an instrument (indeed, the few Greek, Chinese and other foreigners among the contributors show scriptural characteristics apart form the Swiss majority). If the present usage of Rex is rather limited to a browser of a specific dataset and much development can be imagined, it is nevertheless also an intriguing tool to experiment with as a testbed for other computational forensic applications.
it has a number of less apparent properties with implications for
the expertise work. They reveal why studies find orientation not
the best performing biometric instrument [
Rotation — The feature is evidently not rotation invariant, meaning that the same document will have different measurement profiles depending on, for example, the skew of the paper in a scanner (Fig. 5.1–2). However the difference is only a translation of the profile, thus the bias can be corrected.
Organization — Contour orientation exhibits some unusual cases of shape invariance, all deriving from its low sensitivity to the spatial organization of pixels, due to the fact that, by definition, the measure is done locally. It is thus possible to have perceptually different shapes with the same orientation profile. Fig. 5.5 demonstrates scrambling invariance.
Localization — The various informations that can be read in the global orientation profile can’t be traced to specific locations in the written document. If there is, say slant variation in a particular line, we see it in the profile, but can’t localize the given line and even not know if the variation is concentrated in one line or spread over the entire document.
Convexity — For 180° shape rotations the profiles are identical, leading to shape confusion (Fig. 5.3–4).
Neighborhood — Fig. 5.6 shows that lines and circles in certain configurations can look the same to the orientation instrument: it is unaware about the neighborhood.
Additivity — Shapes contribute linearly to profiles, facilitating combinatorial pattern simulations from primitives.