Noise-Induced Training for Weak Signal Detection in Neyman–Pearson Framework

Abstract

Here, we propose a noise-induced neural network-based detector. The suggested method performs well in the detection of the known weak DC signal in additive to non-Gaussian noise. The precalculated noise is added in a neural network which helps in boosting the performance of the weak signal detector. This precalculated noise boosts the training process for signal detection. While training, the backpropagation (BP) algorithm acquires less error and it converges faster with the addition of the external noise. This method performs better than the traditional neural network-based detector in terms of its performance characteristics, i.e., the probability of detection (\(P_D\)) at a fixed value probability of false alarm (\(P_{FA}\)). We also test our noise-induced proposed detector under several signal-to-noise ratio (SNR) environments. The different state-of-the-art techniques have been compared with our proposed method.