SOM-SVM模型是利用SOM的聚类特点，将含有相同特征的输入样本聚集在一起，并把离聚类中心较远的输入样本舍去。经过20%的样本压缩后，将含有代表性的小样本再送入SVM进行训练。本文的样本集通过实验平台采集，验证了基于支持向量机的频谱感知方法在实际数据测试条件下也能取得很好的感知性能。仿真结果表明，SOM-SVM模型在低信噪比下，频谱检测率接近100%，检测错误率也得到了很好的改善。(The SOM-SVM model is based on clustering characteristics of SOM, the input sample will contain the same features together, and the input samples from far down the clustering center. After 20% samples were compressed, the representative samples were sent to SVM for training. In this paper, the sample set is collected through the experimental platform, and the spectrum sensing method based on support vector machine is proved to be able to achieve good perception performance under the actual data testing conditions. The simulation results show that the detection rate of SOM-SVM model is close to 100% at low SNR, and the detection error rate is also improved.)