Research on Nonnegative Matrix Decomposition Spectrum Perception Method Based on Convolutional Neural Networks

In order to improve the spectrum sensing performance of cognitive radio systems under low signal-to-noise ratio conditions, a non negative matrix decomposition spectrum sensing method based on convolutional neural networks is proposed. The core idea of this method is to use convolutional neural network structure to effectively extract spectral features, and combine non negative matrix decomposition technology to decompose and reconstruct the observed signals. Through this approach, noise and signal can be effectively distinguished, thereby improving the system's perception of the surrounding spectral environment. The simulation results show that this method has better perceptual performance than existing methods at a signal-to-noise ratio of -25dB~-5dB.