一种基于统计原理的特征参数提取法

A Feature Extraction Method Based on Statistical Theory

在模拟信号和数字信号的联合制式识别中,通常使用零中心归一化瞬时幅度紧致性参数来区分AM和MASK信号,使用零中心归一化瞬时频率的紧致性参数来区分FM和MFSK信号。但是,在实际的无线电监测环境中,通信信号由于受电磁干扰、设备热噪声等因素的影响,采用这两种紧致性参数提取并识别与之对应的模拟调制与数字键控信号时识别率低,不能满足工程应用。为了提高识别率,文中提出了一种基于平滑统计原理的特征参数提取法,该方法能准确估计信号瞬时幅值或频率在统计空间上的峰值数。工程应用表明,该方法简单可靠,受噪声影响小,识别率高。

In the automatic recognition of analog and digital modulations ,the discrimination between the AM signals and the MASK ones can be achieved by comparing the kurtosis of the normalised-centered instantaneous amplitude ,while The FM and MFSK signals can be distinguished by comparing the kurtosis of the normalised-centered instantaneous frequency. However ,in the practical radio monitoring environment, communication signals are interrupted due to electromagnetic interference and thermal noises from equipments . The recognition rate achieved by using these two parameters to extract and identify the modulation of its corresponding analog and digital signal is rather low, which can not meet application requirements. In order to improve the recognition rate, the paper puts forward a feature extraction method based on statistical theory, which can accurately estimate the number of peaks of signal transient amplitude or frequency in the statistics space. It is proved in application that the method is simple and reliable , less affected by noise and leads to high recognition rate.