中文摘要：

支持向量机算法在解决小样本、非线性及高维模式识别问题中存在特有优势,广泛应用于统计分类及回归分析中。传统的频移键控解码在低信噪比下的误码率较高,解码结果受器件性能的影响较大,因此提出将支持向量机用于频移键控解码。构建分类训练样本点集,通过仿真训练找出合适的核函数,获得解码结果。该方法与相干检测、非相干检测以及基于神经网络的解码方法针对同一信源解码,将不同信噪比下的误码率进行对比分析。实验结果表明,该方法在较低信噪比情况下仍然显示出很好的解码性能和良好的稳健性,优于已有信号解码方法。

英文摘要：

SVM algorithm has unique advantages in solving small sample,nonlinear and high dimensional pattern recognition problems,and it is widely used in statistical classification and regression analysis. Traditional frequency shift keying decoding algorithms have higher bit error rate in low signal to noise ratio,and greater impact on the result of decoding by performance of devices. In this paper,SVM algorithm is applied to decoding frequency shift keying. It constructed sample points set for classification training, and found out the most suitable kernel function through simulation training to obtain the decoding result. The method is compared with coherent detection,non-coherent detection and decoding method based on neural networks for the same source,and analyzes the bit error rate under different SNR. The experimental results show that the method still shows good decoding performance and good robustness at lower SNR,which is superior to the existing signal decoding method.