Spectrum Analysis of 2.4 GHz Band Based on Successive Signal Detection

The high cost of cellular spectrum has motivated network providers to extend LTE to unlicensed band. The 2.4 GHz band is an attractive choice since it provides higher penetration and coverage. However, operating in 2.4 GHz band inevitably suffers serious interference from heterogeneous network devices. We in this paper propose a 2.4 GHz band spectrum analysis method, which is the important first step to build reliable connections among LTE devices in the 2.4 GHz band. Our method involves two stages. It first locates and captures the interference signal on the time-frequency spectrogram with a region growing-based image segmentation method. With the extracted signal, it then identifies the type of interference technology based on Convolutional Neural Networks (CNN). The obtained information (i.e., the type and the time-frequency position of the interference on the spectrogram) provided the base for the interference avoidance and resistance methods. Experimental evaluation shows that the proposed spectrum analysis system can locate low-power (1.09 dB) signal with high accuracy and achieves a classification accuracy of 96.15%.