| 摘要: |
| 针对传统随机森林调制识别算法计算量大且无法应用于高阶调制格式识别的问题，提出了一种基于特征筛选的随机森林调制识别方法。首先，提出了一种通过相位信息构建新信号的方法提取高阶矩谱特征和高阶累积量特征，使随机森林分类器的调制识别准确率能力从低阶FSK、PSK、16QAM和MSK信号扩展到32APSK、64APSK(16+16+16+16)、64APSK(8+16+20+20)、128APSK和256APSK信号。其次，从9种特征中筛选出皮尔逊相关系数较小的信号特征作为随机森林分类器的输入，保留了最具代表性的特征信息的同时进一步减少了计算量。该模型对关注的高级调制格式有着良好的调制识别能力，在信噪比为7 dB时调制识别准确率可达95%以上。 |
| 关键词: 高阶调制识别 随机森林 皮尔逊相关系数 |
| DOI：10.20079/j.issn.1001-893x.240831002 |
|
| 基金项目: |
|
| Random Forest-based Higher-order Modulation Recognition Approach |
| HUANG Guoqiang,FU Jingyi,XU Jing |
| (1.Southwest China Institute of Electronic Technology,Chengdu 610036,China;2.School of Communication and Electronic Engineering,East China Normal University,Shanghai 200062,China) |
| Abstract: |
| For the problem that the traditional random forest modulation identification algorithm is computationally intensive and cannot be applied to the identification of higher-order modulation formats,a random forest modulation identification model based on feature screening is proposed.Firstly,a new signal is constructed by extracting the instantaneous phases of eleven modulation formats,and the higher-order moment spectrum features and higher-order cumulants of these signals are extracted to extend the classification capability of the random forest modulation recognition model from low-order FSK,PSK,16QAM and MSK signals for classification of 32APSK,64APSK(16+16+16+16),64APSK(8+16+20+20),128APSK and 256APSK signals.Secondly,the signal features with smaller Pearson correlation coefficients are selected from the nine features as inputs to the random forest classifier,which retains the most representative feature information while further reducing the computational effort.This model has outstanding classification ability for the high-order modulation formats of interest,and the classification accuracy can reach more than 95% at a signal-to-noise ratio of 7 dB. |
| Key words: higher order modulation recognition random forest Pearson correlation coefficient |
《电讯技术》编辑部 