Abstract
The project involves the channel estimation of the received signal transmitted using the OFDM multiplexing technique. Deep learning is employed and is used in the physical layer in wireless communication system. Deep learning model is trained based on the received input data and output which is the actual data transmitted through OFDM channels. The project aims to improvise the existing model on the same dataset in order to minimize Bit error rate (BER). Different Models like Convolution Neural Network, Bidirectional LSTM and Conv-LSTM are considered with parameter tuning to get better results. A new dataset containing SNR and Modulations is taken in order to classify modulations using CGAN (Conditional Generative Adversarial Networks) and to get better accuracy. In CGANs labels act as a condition to generate and discriminate images better. Modulations have different patterns and CGAN can thus be applied to recognize images and thus the generated images are perceived to be better. This is applied for every SNR to get the results.