**10. References**


260 Bio-Inspired Computational Algorithms and Their Applications

worst case. Compared to a typical implementation with backpropagation neural networks and cepstral coefficients, this approach was at least 10% more effective in near real-life

The genetic algorithm consistently maximized the criteria of inter-class separability and intra-class compactness, under different conditions of population, probability of mutation, etc., and also varying the restriction set. It is remarkable and worth to mention that the genetic algorithm didn't gave the best result in its first execution, which means that the execution must be repeated to achieve good results; increasing the population and manipulating the restriction set demonstrated that it is possible to obtain a variety of

The future work will consist of developing the voice/unvoiced detection in noisy environments, investigate an adapt more features that can be easily computed in time and with a dual in frequency, start working on the non-dependant speaker approach, making use of more robust classifiers, and finally, increase the vocabulary size, although still restricted to a specific semantic field, like in [35]. Another interesting venue is the one in which the user aging process is taken into account by the speech recognition system [36].

The authors wish to express their gratitude for financial support of this project to

[1] K. Koutroumbas and S. Theodoridis, *Pattern Recognition*, 1st ed. California, E. U. A.:

[2] John G. Proakis and Dimitris G. Manolakis, *Digital Signal Processing. Principles, Algorithms, and Applications.*, 3rd ed. New Jersey, U.S.A.: Prentice Hall, 1996. [3] L. R. Rabiner and R. W. Schafer, *Introduction to Digital Signal Processing*, 1st ed.

[4] Isabelle Guyon and André Elisseeff, An Introduction to Variable and Fature Selection,

[5] R. O. Duda, P. E. Hart, and D. G. Stork, *Pattern Classification*, 1st ed. New York, U.S.A.:

[6] D. E. Goldberg, *Genetic Algorithms in Search, Optimization and Machine Learning*. U. S. A.:

[7] F. Itakura, Minimum Prediction Residual Principle applied to Speech Recognition, *IEEE* 

[8] J. Makhoul, Linear Prediction: A Tutorial Review, *Proceedings of IEEE*, vol. 63, no. 4, pp.

[9] C. D. Manning, *Foundations of Statistical Natural Language Processing*, 6th ed. Cambridge,

[10] J. Ramírez, J. M. Górriz, and J. C. Segura, Voice Activity Detection. Fundamentals and

*Transactions on Acoustics, Speech, and Signal Processing*, vol. 23, no. 2, pp. 67-72, 1975.

Speech Recognition System Robustness, in *Robust Speech Recognitin and Understanding*, M. Grimm and K. Kroschel, Eds. Vienna, Austria: In-Tech, 2007,

*Journal of Machine Learning Research*, vol. 3, pp. 1157-1182, 2003.

different outcomes, so it is important to experiment carefully.

Hannover, U.S.A.: Now Publishers Inc., 2007.

application.

**9. Acknowledgment** 

**10. References** 

*Universidad Politécnica de Aguascalientes*.

Academic Press, 1999.

John Wiley & Sons, Inc., 2001.

561-580, 1975.

cap. 1, pp. 1-22.

Addison-Wesley Professional, 1989.

Massachussets, U.S.A.: MIT Press, 2003.


**13** 

*Malaysia* 

**Performance of Varying Genetic** 

 **Algorithm Techniques in Online Auction** 

Kim Soon Gan, Patricia Anthony, Jason Teo and Kim On Chin

*Universiti Malaysia Sabah, School of Engineering and Information Technology, Sabah* 

Genetic algorithm is one of the successful optimization algorithm used in computing to find exact or approximate solutions for certain complex problems. This novel algorithm was first introduced by John Holland in 1975 (Holland, 1975). Besides Holland, many other researchers have also contributed to genetic algorithm (Davis, 1987; Davis, 1991; Grefenstte, 1986; Goldberg, 1989; Michalewicz, 1992). This is an algorithm that imitates the evolutionary process concept based on the Darwinian Theory which emphasizes on the law of "the survival of the fittest". This algorithm used techniques which are inspired from evolution biology such as inheritance, selection, crossover and mutation

There are several important components in genetic algorithm which includes representation, fitness function, and selection operators (parent selection and survivor selection, crossover operator and mutation operator). Genetic algorithm starts by generating an initial population of individuals randomly. The individuals are represented as a set of parameter which is the solution to the problem domain. Normally, individuals are fixed length binary string. The individuals are then evaluated using fitness functions. The evaluation will give a fitness score to individuals indicating how well the solutions perform in the problem domain. The individuals that have been evaluated using the fitness function will be selected to be parents to produce offspring through the crossover and mutation operators. The genetic algorithms will repeat the above process except for the population initialization until the termination criteria is met. Fig. 1 shows the structure of a genetic

GAs have been applied successfully in many applications including job shop scheduling (Uckun *et al.* 1993), the automated design of fuzzy logic controllers and systems (Karr 1991; Lee & Takagi, 1993), hardware-software co-design and VLSI design (Catania *et al.* 1997; Chandrasekharam *et al.* 1993). In this chapter, variations of genetic algorithms are applied in

Auction is defined as a bidding mechanism and is expressed by a set of auction rules that specify how the winner is determined and how much he or she has to pay (Wolfstetter, 2002). Jansen defines an online auction as an Internet-based version of a traditional auction (Jansen, 2003). In today's e-commerce market, online auction has acted as an important tool

optimizing the bidding strategies for a dynamic online auctions environment.

**1. Introduction** 

(Engelbrecht, 2002).

algorithm.

Conversational Telephone Speech, in *IEEE Instrumentation and Measurement Technology Conference*, Budapest, Hungary, 2001, pp. 1926-1931.

