Precoder Design for K-User Interference Channels with Finite Alphabet Signals

We consider a MIMO K-user interference channel with finite alphabet signaling, and design an optimal precoder for maximizing the sum mutual information between all transmitter-receiver pairs. Perfect channel knowledge is assumed to be available at all the transmitters, but, decoding of the interfering signals is not done at any receiver. A gradient descent algorithm is employed for maximizing the sum mutual information. Simulation results for the three user interference channel show that when compared to conventional schemes which maximize the signal-to-interference plus noise ratio or minimize the mean squared error, the proposed design achieves a significantly better bit error rate performance. Along with the proposed precoder, the minimum distance detector which decodes the desired signal (while treating interference as colored Gaussian noise with a known covariance) was utilized at all the receivers, in order to get this enhanced performance.