Acoustics developed intelligent constant false alarm detector based on support vector machine technology

[Research and Development of China Instrument Network Instrumentation] Recently, Wang Leou, Assistant Researcher of the Underwater Vehicle Information Technology Key Laboratory of the Institute of Acoustics, Chinese Academy of Sciences, and his collaborators have used Support Vector Machine (SVM) technology in machine learning. An intelligent Constant False Alarm Rate (CFAR) detector.

When the background noise is unknown, the CFAR detector is a very useful method for adaptive radar detection. However, traditional CFAR detection methods, such as mean detectors, ordered statistic detectors, and adaptive detectors, are difficult to control both detection and false alarm performance under uniform background and non-uniform background.

The CFAR detector proposed in this study uses prior data to train the SVM, then uses the trained SVM to identify the current working environment and outputs a judgment signal, and intelligently selects an appropriate detection threshold according to the judgment signal. It can provide optimal detection performance under a uniform background environment and improve the robustness of detection performance in a non-uniform background environment. After testing, the detection performance of the detector in different environments is superior to the traditional method. The study successfully transplanted machine learning technology into the traditional signal detector field, providing new ideas for signal detection in non-Gaussian contexts.

Related research results were published on IEEE Access.

(Original title: Acoustics based on support vector machine technology developed intelligent constant false alarm detector)