-
[show abstract]
[hide abstract]
ABSTRACT: In this paper, an efficient foreground/background video coding
algorithm is proposed. The algorithm consists of three major parts: an
automatic human face detection and facial feature extraction, facial
feature tracking, and video coding method based on the properties of the
facial feature. In the first part, the location of the human face and
its respective facial features are extracted by means of a genetic
algorithm and eigenface techniques. After the face region and the facial
features have been detected in the first frame, a fast and robust method
is proposed for tracking the detected facial features on the subsequent
frames. Finally, the location of the facial features are passed to the
encoding system for compression. As a result, both the subjective and
the objective image quality of the face region are improved
Intelligent Multimedia, Video and Speech Processing, 2001. Proceedings of 2001 International Symposium on; 02/2001
-
[show abstract]
[hide abstract]
ABSTRACT: A framework to automatically generate a reduced-order
discrete-time model for the sampled system of a continuous plant
preceded by a zero-order hold (ZOH) using an enhanced multi-resolution
dynamic genetic algorithm (EMDGA) is proposed in this paper. Chromosomes
consisting of the denominator and the numerator parameters of the
reduced-order model are coded as a vector with floating-point-type
components and searched by the genetic algorithm. Therefore, a stable
optimal reduced-order model satisfying the error range specified can be
evolutionarily obtained. Because of the use of the multi-resolution
dynamic adaptation algorithm and the genetic operators, the convergence
rate of the evolution process to search for an optimal reduced-order
model can be expedited. Another advantage of this approach is that the
reduced discrete-time model evolves based on samples taken directly from
the continuous plant, instead of the exact discrete-time model, so that
computation time is saved
Systems, Man, and Cybernetics, 2001 IEEE International Conference on; 02/2001
-
[show abstract]
[hide abstract]
ABSTRACT: The conventional approach to find the constrained minimum-time
path for robot manipulator employs the trial-and-error procedure, namely
the flexible polyhedron search method. In this paper we introduce an
alternative approach by applying the genetic search algorithms to
schedule the time intervals between each pair of adjacent knots such
that the total travelling time is minimized subjected to the physical
constraints on joint velocities, accelerations, and jerks. Modified
heuristic crossover and a scaled and normed performance measure are
applied to the genetic algorithmic searching procedures. Experiments
with different combinations of crossover rates and mutation rates are
carried out and the corresponding results outweigh the constrained
minimum-time obtained from the trial-and-error polyhedron search method
Systems, Man, and Cybernetics, 1998. 1998 IEEE International Conference on; 11/1998
-
[show abstract]
[hide abstract]
ABSTRACT: A genetic-algorithms-based approach which takes the closed-loop
system into account to digitally redesign an analogue control system is
proposed. In comparison with conventional discretisation methods, the
performance of the redesigned system using the resulting digital
controller is significantly improved, while the drawback of excessive
order of the controller is avoided
Electronics Letters 08/1997; · 0.96 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Investigations into data storage schemes for parallel memory
system of vector processing have been mainly focused on low-order
interleaved schemes, skewed schemes and XOR schemes. In this paper, a
new interleaved storage scheme, namely k-row interleaved scheme, is
suggested and investigated. This scheme allocates k consecutive data of
a vector onto one memory module sequentially and then the next k
consecutive data onto the next memory module. The address mapping
functions are devised and the performance of this scheme is evaluated.
It is found that this scheme improves the average performance for vector
access over the previous schemes. The address generation hardware is
also shown to be simple
TENCON '93. Proceedings. Computer, Communication, Control and Power Engineering.1993 IEEE Region 10 Conference on; 11/1993
-
01/1993
