A Novel WLAN Receiver Performance in Highly Dispersive and Nonlinear Environment Shubhangi Rathkanthiwar 1 , Chandrashekhar Dethe 2 and Kishore Kulat 1+ 1 Associate Professor, Yeshwantrao Chavan college of Engineering, Nagpur, India 2 Principal, Priyadarshini Institute of Technology, Nagpur, India 3 Professor and Head, Department of Electronics Engg, Viswesaraya National Institute of Technology, Nagpur, India Abstract: An efficient, adaptive and intelligent transmission–reception system, which will take care of challenging fading multipath problems in real case scenario, is always needed in digital communications. This paper presents a novel, neural based method to improve performance of WLAN receiver in presence of nonlinear distortions introduced by high power amplifiers, Doppler effect, delay spread and many other fading multipath problems. Keywords: BER, HPA, OFDM, SOM, PAPR, WLAN 1. Introduction There is still a gap between previous research conducted on performance improvement in receiver structure in fading multipath environment and use of neural networks as learning systems which can incorporate constraints on their capability to handle several problems coming in upcoming technologies in the wireless communications [1]. Main gap is exponentially growing demand for great quality services at high data rates and implementation of structures using computational intelligence to take care of many complex cases arising in multicarrier communications [2,3]. As OFDM is the most promising candidate in present as well as future generation wireless communications, we have selected wireless local area network model Hiperlan2 for constructing this paper, as it uses OFDM as multiplexing/ modulation technique. One of the most challenging issues in OFDM based system that has still remained unresolved is the problem of nonlinear distortion. It has been taken into consideration along with multipath fading problems. Main idea behind use of neural networks in OFDM based systems is the signal classification. If the classification is done properly in presence of environment harshly deteriorating the signal, we say that there is an improvement in the performance of the system. Neural networks are the learning systems that allow the people to specify what the systems should do for each case. It can decide in a reasonable way what to do in a particular situation from previous experiences and/or provided examples of appropriate behaviors even though the situation may not be experienced by the system before. Review of research work clearly indicated limitations of the supervised network approaches like MLP and RBF in handling migratory signals whose stochastic properties such as average values of signals in each cluster are varying continuously. We have identified the self-organizing map (SOM) method as a powerful software tool for the visualization of high-dimensional data. It converts complex, nonlinear statistical relationships between high dimensional data into simple geometric relationships on a low dimensional display. It thereby compresses information while preserving the most important topological relationships of the primary data elements. Visualization and abstraction are the two aspects that occur in a number of complex engineering tasks such as process analysis, machine perception, control, and communication[4]. 2012 International Conference on Networks and Information (ICNI 2012) IPCSIT vol. 57 (2012) © (2012) IACSIT Press, Singapore DOI: 10.7763/IPCSIT.2012.V57.01 1