Open AccessDetection of weak monocycle sinusoidal signals with a low constant false alarm rate based on the support vector machine
Author(s) -
Tan Bo,
Guo Jingbo,
Chang Guang
Publication year - 2019
Publication title -
the journal of engineering
Language(s) - English
Resource type - Journals
ISSN - 2051-3305
DOI - 10.1049/joe.2018.8584
Subject(s) - false alarm , constant false alarm rate , noise (video) , white noise , signal (programming language) , support vector machine , detection theory , additive white gaussian noise , signal to noise ratio (imaging) , computer science , constant (computer programming) , pattern recognition (psychology) , gaussian noise , alarm , statistical power , speech recognition , statistics , artificial intelligence , mathematics , telecommunications , engineering , electrical engineering , programming language , detector , image (mathematics)
The detection of weak monocycle sinusoidal signals is the abstract problem of the weak transient signal detection in many engineering applications. A low constant false alarm rate is often required on some occasions with serious false alarm consequences, such as the relay protection and the weapon trigger. The weak monocycle sinusoidal signal means a low signal‐to‐noise ratio (SNR), only one sinusoidal period, and an unknown initial phase. The support vector machine (SVM) is applied to detect the weak monocycle sinusoidal signal. The false alarm performance of the built SVM is attractive. The characteristics of the built SVM in white Gaussian noise are as follows. Firstly, the probability of false alarm is 0 and the probability of detection is 1 when the SNR is over −11 dB. Secondly, different unknown initial phases cannot influence the performance. Thirdly, the probability of false alarm is always 0, when the SNR is over −25 dB. Furthermore, the experimental results show that this method is also effective in coloured noise. When the SNR of coloured noise is −0.28 dB, the weak monocycle sinusoidal signal can be detected. When the SNR of coloured noise is over −60 dB, the probability of false alarm is always 0.