Digital Self-interference Cancellation with Support Vector Regression for Full Duplex Communication

Full duplex (FD) communication is one of the prominent techniques due to its potential to double throughput without increasing the used bandwidth. To enable FD communication, the self-interference (SI) signal at the transmitter should be reduced down to the noise level. The current solutions are not able to cancel SI at all power levels, especially at high power levels. In this paper, a new nonlinear digital cancellation (DC) approach is proposed by adapting support vector regression (SVR) for FD communication. The digital SI cancellation algorithms are tested on a software defined radio set-up with a single antenna. For high transmit power levels, with the proposed SVR-based solution, up to 5 dB higher total cancellation is observed in comparison to linear DC, and up to 3 dB improvement is obtained over the memory polynomial based nonlinear DC. This performance enhancement is provided by implementing the algorithm in the baseband, it does not require any additional hardware and it does not cause any extra communication overhead.