@InProceedings{ba_pe_ba_gi_iv:08,
author = {Claus Bahlmann and Martin Pellkofer and
Jan Giebel and Gregory Baratoff},
title = {Multi-Modal Speed Limit Assistants: Combining Camera and GPS Maps},
booktitle = {IEEE Intelligent Vehicles Symposium (IV 2008)},
year = 2008,
}
@InProceedings{ke_sp_ba_ba_gi_iv:08,
author = {Christoph Gustav Keller and Christoph Sprunk and
Claus Bahlmann and Jan Giebel and Gregory Baratoff},
title = {Real-Time Recognition of U.S. Speed Signs},
booktitle = {IEEE Intelligent Vehicles Symposium (IV 2008)},
year = 2008,
}
@Article{ba_pr:06,
author = {Claus Bahlmann},
title = {Directional Features in Online Handwriting
Recognition},
abstract = {The selection of valuable features is crucial in
pattern recognition. In this paper we deal with the
issue that part of features originate from
directional instead of common linear data. Both for
directional and linear data a theory for a
statistical modeling exists. However, none of these
theories gives an integrated solution to problems,
where linear and directional variables are to be
combined in a single, multivariate probability
density function. We describe a general approach for
a unified statistical modeling, given the constraint
that variances of the circular variables are
small. The method is practically evaluated in the
context of our online handwriting recognition system
frog on hand and the so-called tangent slope angle
feature. Recognition results are compared with two
alternative modeling approaches. The proposed
solution gives significant improvements in
recognition accuracy, computational speed and memory
requirements.},
journal = {Pattern Recognition},
month = jan,
number = 1,
volume = 39,
year = 2006
}
@InProceedings{ba_zh_ra_pe_ko_iv:05,
author = {Claus Bahlmann and Ying Zhu and Visvanathan Ramesh
and Martin Pellkofer and Thorsten Koehler},
title = {A System for Traffic Sign Detection, Tracking, and
Recognition Using Color, Shape, and Motion
Information},
abstract = {This paper describes a computer vision based system
for real-time robust traffic sign detection,
tracking, and recognition. Such a framework is of
major interest for driver assistance in an
intelligent automotive cockpit environment. The
proposed approach consists of two components. First,
signs are detected using a set of Haar wavelet
features obtained from Ada- Boost training. Compared
to previously published approaches, our solution
offers a generic, joint modeling of color and shape
information without the need of tuning free
parameters. Once detected, objects are efficiently
tracked within a temporal information propagation
framework. Second, classification is performed using
Bayesian generative modeling.Making use of the
tracking information, hypotheses are fused over
multiple frames. Experiments show high detection and
recognition accuracy and a frame rate of
approximately 10 frames per second on a standard
PC.},
booktitle = {IEEE Intelligent Vehicles Symposium (IV 2005)},
year = 2005,
}
@PhdThesis{ba_phd:05,
author = {Claus Bahlmann},
title = {Advanced Sequence Classification Techniques Applied
to Online Handwriting Recognition},
abstract = {The term handwriting recognition (HWR) denotes the
process of transforming a language, which is
represented in its spatial form of graphical marks,
into its symbolic representation. Online HWR
performs this task concurrently to the writing
process. The present thesis studies highaccuracy
recognition methods applied to online HWR. Those
methods have been implemented within the writer
independent online HWR system frog on hand (Freiburg
RecOGnition of ON-line HANDwriting). In online HWR,
data are typically represented as vector
sequences. In addition to HWR, vector sequence data
appear in a number of additional pattern recognition
problems, for instance, in speech recognition,
genome processing, financial and medical
applications, and robotics. For those problems,
designing classifiers that directly address the
data's natural representation can greatly improve
the recognition accuracy, compared to a potential
pre-applied transformation to vector space
data. Beside introducing novel online HWR
approaches, a concern of this thesis is also to
develop broadly applicable pattern recognition
techniques, which are generic to this bouquet of
sequence data problems. Emphasis is placed on
classification. This thesis describes two
complementary classification methods, one of them
(CSDTW) falling into the so-called generative, the
other one (SVM-GDTW) into the so-called
discriminative classification category. The
generative CSDTW (cluster generative statistical
dynamic time warping) is a scalable sequence
classification, which aims at holistically combining
sequence cluster analysis and statistical
modeling. Contrary to previous approaches, these two
aspects are embedded in a single feature space and
use a closely related distance measure. As will be
shown, this combined modeling leads to very accurate
HWR results. Particularly interesting in the context
of statistical classification, like CSDTW, is the
modeling of so-called directional data (i.e., data
which corresponds to a direction, thus, in 2D is
distributed on the unit circle; opposed to
directional data, linear data is distributed along
the real line). In online HWR directional data
appear as a valuable feature by means of the angular
pen trace direction. This thesis describes a unified
modeling of directional and linear data within one
probability density function (PDF): the multivariate
semi-wrapped Gaussian PDF. This modeling applied to
CSDTW classification shows significant improvements
in recognition accuracy, computational speed and
memory requirements, compared to commonly employed
modeling approaches. As an additional resource for
the CSDTW sequence modeling, a (dis-) similarity
measure between a pair of CSDTW models is
described. Such a measure can be used as a stop
criterion in the iterative CSDTW training, as a
speed-up in classification, a distance measure in
the context of CSDTW model clustering or as an
optimization criterion for a discriminative CSDTW
training. Likewise to the CSDTW scoring, this (dis-)
similarity computation uses dynamic programming as
algorithmic framework and can thus be easily added
to a given classification implementation. It is
based on the Bayes probability of error, and, hence,
can be utilized as a tool to interpret
misclassifications. Experiments show a high
correlation of similar and frequently confused class
pairs. As a complementary approach to the widely
employed generative sequence modeling, a
discriminative strategy of fusing dynamic time
warping (DTW) and support vector machines (SVM) is
developed: SVM-GDTW. This fusion is realized by a
formulation of a novel SVM kernel, called the
Gaussian dynamic time warping (GDTW) kernel. As this
sequence classification approach is a pure
discriminative one, it does not assume a model for
the generative class conditional densities. Instead,
it addresses the direct creation of class
boundaries. This thesis compares CSDTW and SVM-GDTW
in terms of theoretical background, accuracy, and
computational complexity. While CSDTW being the more
efficient approach, SVM-GDTW holds much potential
for future research as an instance of the relatively
recent SVM based sequence classification. The
practical impact of the developed handwritten
character recognition is demonstrated by an
implementation on a Linux Compaq iPAQ PDA
environment.},
address = {Institut f\"ur Informatik},
institution = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
school = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
publisher = {Shaker-Verlag},
year = {2005},
url = {http://www.shaker.de/Online-Gesamtkatalog/Details.asp?ID=837419&CC=17185&ISBN=3-8322-4535-9&Reihe=15&ReiheUR=-2&start=1},
isbn = {3-8322-4535-9}
}
@InProceedings{ha_ba_dagm:04,
author = {Bernard Haasdonk and Claus Bahlmann},
title = {Learning with Distance Substitution Kernels},
booktitle = {26th Pattern Recognition Symposium of the German
Association for Pattern Recognition (DAGM 2004)},
abstract = {During recent years much effort has been spent in
incorporating problem specific a-priori knowledge
into kernel methods for machine learning. A common
example is a-priori knowledge given by a distance
measure between objects. A simple but effective
approach for kernel construction consists of
substituting the Euclidean distance in ordinary
kernel functions by the problem specific distance
measure. We formalize this distance substitution
procedure and investigate theoretical and empirical
effects. In particular we state criteria for
definiteness of the resulting kernels. We
demonstrate the wide applicability by solving
several classification tasks with
SVMs. Regularization of the kernel matrices can
additionally increase the recognition accuracy.},
year = 2004,
address = {T\"ubingen, Germany},
publisher = {Springer Verlag}
}
@Article{ba:bu:tpami04,
author = {Claus Bahlmann and Hans Burkhardt},
title = {The Writer Independent Online Handwriting
Recognition System \emph{frog on hand} and Cluster
Generative Statistical Dynamic Time Warping},
journal = {IEEE Trans. Pattern Anal. and Mach. Intell.},
month = mar,
year = 2004,
volume = 26,
number = 3,
pages = {299--310},
abstract = {In this paper, we give a comprehensive description
of our writer-independent online handwriting
recognition system frog on hand. The focus of this
work concerns the presentation of the
classification/training approach, which we call
cluster generative statistical dynamic time warping
(CSDTW). CSDTW is a general, scalable, HMM-based
method for variable-sized, sequential data that
holistically combines cluster analysis and
statistical sequence modeling. It can handle general
classification problems that rely on this sequential
type of data, e.g., speech recognition, genome
processing, robotics, etc. Contrary to previous
attempts, clustering and statistical sequence
modeling are embedded in a single feature space and
use a closely related distance measure. We show
character recognition experiments of frog on hand
using CSDTW on the UNIPEN online handwriting
database. The recognition accuracy is significantly
higher than reported results of other handwriting
recognition systems. Finally, we describe the
real-time implementation of frog on hand on a Linux
Compaq iPAQ embedded device.},
url =
{ftp://ftp.informatik.uni-freiburg.de/papers/lmb/ba_bu_tpami04.pdf},
keywords = {Pattern recognition, handwriting analysis, Markov
processes, dynamic programming, clustering},
}
@InProceedings{ba:ha:bu:iwfhr02,
author = {Claus Bahlmann and Bernard Haasdonk and Hans
Burkhardt},
title = {On-line Handwriting Recognition with Support Vector
Machines---A Kernel Approach},
pages = {49--54},
booktitle = {Proc. 8th Int. Workshop Front. Handwriting
Recognition (IWFHR)},
year = 2002,
url =
{ftp://ftp.informatik.uni-freiburg.de/papers/lmb/ba_ha_bu_iwfhr02.pdf},
abstract = {In this contribution we describe a novel
classification approach for on-line handwriting
recognition. The technique combines dynamic time
warping (DTW) and support vector machines (SVMs) by
establishing a new SVM kernel. We call this kernel
\emph{Gaussian DTW (GDTW) kernel}. This kernel
approach has a main advantage over common HMM
techniques. It does not assume a model for the
generative class conditional densities. Instead, it
directly addresses the problem of discrimination by
creating class boundaries and thus is less sensitive
to modeling assumptions. By incorporating DTW in the
kernel function, general classification problems
with variable-sized sequential data can be
handled. In this respect the proposed method can be
straightforwardly applied to all classification
problems, where DTW gives a reasonable distance
measure, e.g.~speech recognition or genome
processing. We show experiments with this kernel
approach on the UNIPEN handwriting data, achieving
results comparable to an HMM-based technique. }
}
@InProceedings{ba:bu:icdar01,
author = {Claus Bahlmann and Hans Burkhardt},
title = {Measuring {HMM} Similarity with the {B}ayes
Probability of Error and its Application to Online
Handwriting Recognition},
booktitle = {Proc. 6th Int. Conf. Doc. Anal. Recognition (ICDAR)},
year = 2001,
abstract = {We propose a novel similarity measure for Hidden
Markov Models (HMMs). This measure calculates the
Bayes probability of error for HMM state
correspondences and propagates it along the Viterbi
path in a similar way to the HMM Viterbi scoring. It
can be applied as a tool to interpret
misclassifications, as a stop criterion in iterative
HMM training or as a distance measure for HMM
clustering. The similarity measure is evaluated in
the context of online handwriting recognition on
lower case character models which have been trained
from the UNIPEN database. We compare the
similarities with experimental classifications. The
results show that similar and misclassified class
pairs are highly correlated. The measure is not
limited to handwriting recognition, but can be used
in other applications that use HMM based methods. },
pages = {406--411},
url = {ftp://ftp.informatik.uni-freiburg.de/papers/lmb/ba_bu_icdar01.pdf}
}
@Article{ba:he:ri:pr99,
author = {Claus Bahlmann and Gunther Heidemann and Helge
Ritter},
title = {Artificial Neural Networks for Automated Quality
Control of Textile Seams},
journal = {Pattern Recognition},
volume = 32,
month = jun,
abstract = {We present a method for an automated quality control
of textile seams, which is aimed to establish a
standardized quality measure and to lower costs in
manufacturing. The system consists of a suitable
image acquisition setup, an algorithm for locating
the seam, a feature extraction stage and a neural
network of the self-organizing map type for feature
classification. A procedure to select an optimized
feature set carrying the information relevant for
classification is described.},
url =
{ftp://ftp.informatik.uni-freiburg.de/papers/lmb/ba_he_ri_pr99.ps.gz},
keywords = {neural networks, self-organizing feature maps
(SOFM), textile seams, quality control, feature
selection },
number = 6,
pages = {1049--1060},
year = 1999
}
@Mastersthesis{bahlmann:97,
title = {{K\"unstliche Neuronale Netze zur optischen
Qualit\"atskontrolle textiler N\"ahte}},
author = {Claus Bahlmann},
month = may,
year = 1997,
school = {Universit{\"a}t Bielefeld},
address = {Technische Fakult{\"a}t, AG Neuroinformatik},
url =
{http://lmb.informatik.uni-freiburg.de/people/bahlmann/data/bahlmann-seams1997.pdf},
}
@MastersThesis{simon:03,
author = {Kai Simon},
title = {{E}rkennung von handgeschriebenen {W}\"ortern mit
{CSDTW}},
abstract = {Am LMB werden seit mehreren Jahren Untersuchungen
zur on-line Zeichenerkennung durchgeührt. Ein
System, welches innerhalb dieser Forschungsarbeiten
entwickelt werden konnte ist das CSDTW (Cluster
generative Statistical Dynamic Time Warping). CSDTW
grü ndet auf einer clusterbasierten, generativen,
statistischen Modellierung der Schriftzeichen,
welche mittels DTW-Verfahren zur Klassifikation
eines unbekannten Schriftzeichens herangezogen
werden. Diesbezüglich kommt eine maximum-aposteriori
(MAP) bzw. maximum-likelihood (ML) Klassifikation
zum Einsatz, in der die
Produktionswahrscheinlichkeit für die in Frage
kommenden statistischen Modelle maximiert
wird. Mithilfe diese Systems konnten Fehlerraten von
etwa 10\% für isoliert geschriebene Kleinbuchstaben,
auf dem in der Handschriftforschung
gebräuchlichen "UNIPEN"-Benchmark erreicht
werden. Dies ist ein im internationalen Vergleich
hervorragende Quote. Der vorliegende Vortrag
beschäftigt sich mit der Erweiterung des Systems
im Hinblick auf die Worterkennung. Hierzu werden
Vorverarbeitungstechniken vorgestellt, die eine
Merkmalsextraktion anhand von geschriebenen
Wörten ermöglichen. Schliefllich erfolgt eine
Erweiterung von CSDTW zu Erkennung von
Zeichensequenzen. Zur Einschränkung des Suchraums
kommen Viterbi- bzw. Strahlsuchstrategien und
Pruningtechniken, wie sie im Zusammenhang mit
"Hidden Markov Modellen" (HMMs) bekannt sind, zum
Einsatz. Zusätzlich wird linguistisches
Kontextwissen der zu modellierenden Sprachen in Form
eines Lexikon hinzugezogen. Zum Schluss werden die
erzielten Klassifikationsraten auf dem UNIPEN
Datensatz vorgestellt und diskutiert.},
year = 2003,
language = german,
address = {Institut f\"ur Informatik},
url = {http://lmb.informatik.uni-freiburg.de/},
school = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
}
@MastersThesis{simon:02,
author = {Kai Simon},
title = {{V}orverarbeitung und {M}erkmalsextraktion in der
{O}n\-line-{H}andschrifterkennung},
abstract = {Für eine erfolgreiche Klassifikation eines
Schriftzuges ist die Gewinnung von geeigneten,
gegenüber Translation, Rotation und Skalierung
invarianter Merkmale notwendig. Damit eine robuste
Merkmalsextraktion erfolgen kann, werden die diskret
abgetasteten Ortskoordinaten zuvor mithilfe einer
Splinekurve geglättet. Bei der Glättung kommen
B-Splines zur Anwendung, welche es ermöglichen,
Spitzen in den Abtastpunkten durch Knicke zu
modellieren. Somit bleiben auch nach einer Glättung
wichtige Charakteristiken der Abtastdate n erhalten
und können in eine spätere Merkmalsberechnung
einfließen. Nach dem Erzeugen der Splinekurve,
erfolgt eine Neuabtastung. Hierbei sollen die neu
abgetasteten Punkte äquidistant bezüglich der Kurve
sein. Um dies zu erreichen, wird die Splinekurve
approximativ nach Bogenlänge umparametrisiert. Die
Umparametrisierung stellt einen hohen Rechenaufwand
dar, welcher bei der Online-Handschrifterkennung
nicht akzeptabel ist. Es wird eine schnellere
Alternative vorgestellt, welche jedoch eine größere
Ungenauigkeit in der Äquidistanz der neu
abgetasteten Punkte zur Folge hat. Im Anschluss
werden die invarianten Merkmale 'signed ratio of
tangents' und 'normalized curvature' für die neu
abgetasteten Punkte berechnet. Diese Berechnungen
erfolgen direkt auf Grundlage der Splinedarstellung,
wodurch, im Gegensatz zu einer Interpolation, eine
höhere Rechengenauigkeit gewährleistet ist.},
language = german,
address = {Institut f\"ur Informatik},
url = {http://lmb.informatik.uni-freiburg.de/},
school = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
type = {Student's Thesis},
year = 2002,
}
@MastersThesis{bockhorn:00,
author = {Dirk Bockhorn},
title = {{B}estimmung und {U}ntersuchung von
{S}ignifikanzgewichtungen f\"ur die {E}rkennung von
handgeschriebenen {B}uchstaben},
language = {german},
address = {Institut f\"ur Informatik},
url = {http://lmb.informatik.uni-freiburg.de/},
school = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
year = 2000,
type = {Diploma thesis},
}
@MastersThesis{triebel:99,
author = {Rudolph Triebel},
title = {Automatische Erkennung von handgeschriebenen Worten
mithilfe des Level Building Algorithmus},
address = {Institut f\"ur Informatik},
url = {http://lmb.informatik.uni-freiburg.de/},
school = {Al\-bert-Lud\-wigs-Uni\-ver\-sit\"at Frei\-burg},
year = 1999,
month = dec,
language = {german}
}
|
|
|
