Estimation of indoor infrared channel parameters using neural networks

The indoor infrared channel is modeled as a linear baseband channel. The parameters of this channel are used for the design of communication systems using infrared light as the carrier of information for indoor wireless communications. For many applications, we need to know the spatial distribution of channel parameters such as the received optical power and the delay spread. The simulation process that yields those results is a very time consuming process. We propose using the simulation software to train a neural network using a fraction of the required points and then use the neural network to generate the desired parameters. The results presented show the excellent capability of the neural networks to mimic the simulation software for the purpose of generating the desired set of parameters for an indoor infrared channel. The process is much faster than the simulation software and proper use of sampling set selection yields highly accurate results from the neural network