Proceedings of the Optical / Wireless Communications Technical Symposium, 17 Sep. 2003, Warwick University, UK An Experimental Receiver Architecture For Diffuse Indoor Infrared Environments R J Dickenson and Z Ghassemlooy Optical Communications Research Group, School of Engineering, Sheffield Hallam University, Pond Street, Sheffield, S1 1WB. U.K. Robert J.Dickenson@student.shu.ac.uk, Z.F.Ghassemlooy@shu.ac.uk Web site: http://www.shu.ac.uk/ocr ABSTRACT Wireless networks allow data to be exchanged without the need for fixed connection. Such networks are commonly associated with Radio Frequency (RF) techniques in outdoor environments. More recently, with the advent and widespread use of bluetooth devices and wireless Local Area Network (LAN) products, wireless systems have become more common in the home. It is well understood that the potential data rate offered by the infrared (IR) medium is greater than that offered by current RF technology, however, this high data rate remains untapped in a widespread commercial sense primarily due to the limiting effects of inter symbol interference (ISI) and noise. We propose an alternative to traditional receiver architecture based on the twin concepts of wavelet analysis and artificial intelligence in an attempt to unlock the potentially superior data rates of the IR medium. Keyworks: Diffuse infrared, wavelet analysis, artificial intelligence, inter symbol interference. 1. INTRODUCTION The desirable properties of diffuse IR systems are well published: security, unregulated re- usable bandwidth in adjacent rooms, potential high data rates and so on. However, such systems suffer a similar performance-limiting fate to their RF cousins in terms of noise and ISI; ISI being caused by the many possible paths the original signal can take from the transmitter to the receiver, reflecting off various objects in the room as it does so. For an On Off Keying (OOK) system these multiple paths cause a smearing of the signal over adjacent pulse intervals. It is arguably the ISI that is the most problematic performance degrading effect to overcome in such systems. Further limitations, such as a restricted average transmission power due to ocular safety considerations exacerbate the problem. However, some indoor IR channel characteristics are more favourable than that of RF counterparts. Firstly, diffuse IR systems do not suffer from the affects of fade like an RF system, as the dimensions of the IR wavelength are much smaller in comparison to the size of the optical receiver. Further, the channel is quasi-static, for a 100MHz modulating signal an object would have to move about 3 metres before the characteristics of the channel change [1]. Such qualities may allow us to employ more unconventional receiver designs that provide an opportunity to exploit the potential bandwidth and improve the Bit Error Rate (BER). 2. THE CHANNEL MODEL All our results are based on simulations and models constructed and executed in Matlab. The channel model adopted in this research is from the well-cited paper by Carruthers and Kahn [2], where the channel can be characterised by the normalised delay spread. This dimensionless parameter is defined as the R.M.S delay spread divided by the bit duration. Where the R.M.S delay spread is given as: (1) where a is related the time taken for a once reflected signal to arrive back at the detector, and h(t,a) is the multipath impulse response, given as: (2) where u(t) is the unit step function. 11 13 12 )) , ( ( a a t h D = ( ) ) ( 6 ) , ( 7 6 t u a t a a t h + =