CNN-based blind SIR classification framework for STPA-BAA spectrum superposing

Abstract

Spectrum sharing in the spatial domain among the plurality of wireless communication systems should be addressed due to its exhaustion. We previously conceptualized a spectrum superposing enabled by subcarrier transmission power assignment (STPA) and blind adaptive array (BAA), where co-channel interference can be suppressed even without a priori information of interferer. However, it requires knowledge of input signal-to-interference power ratio (SIR) to realize completely blind operation. This paper proposes a blind SIR estimation framework by convolutional neural network (CNN) using power spectrum images. Simulation verifies the proposed scheme can maximize throughput performance based on SIR classification results with 97% accuracy.

Keywords

• Convolutional neural network
• Blind adaptive array
• Subcarrier transmission power assignment
• Spectrum superposing