CAJ | 학술논문

针对孤立脉冲群电磁超声系统信号特征容易被噪声淹没的问题，提出基于改进的非负矩阵分解（INMF）优选特征的支持向量机（SVM）方法。首先，用3种不同的方法提取高维特征；其次，用NMF方法实现特征降维，并保证降维结果的唯一性，避免对特征的直接选择；最后，应用支持向量机方法对降维特征进行分类。对孤立脉冲群电磁超声系统采集的4种信号特征进行提取、选择和分类，实验结果表明：INMF方法能有效提取微弱信号的特征，减少运算量，提高电磁超声系统特征采集的准确率。
침대고립맥충군전자초성계통신호특정용역피조성엄몰적문제，제출기우개진적비부구진분해（INMF）우선특정적지지향량궤（SVM）방법。수선，용3충불동적방법제취고유특정；기차，용NMF방법실현특정강유，병보증강유결과적유일성，피면대특정적직접선택；최후，응용지지향량궤방법대강유특정진행분류。대고립맥충군전자초성계통채집적4충신호특정진행제취、선택화분류，실험결과표명：INMF방법능유효제취미약신호적특정，감소운산량，제고전자초성계통특정채집적준학솔。
A characteristics optimized method based on improved non-negative matrix factorization(NMF)and support vector machine (SVM)is proposed to solve the problem that the characteris-tic in the signal of isolated impulse cluster electromagnetic acoustic system is easily overwhelmedby the noise.First,high-dimensional feature is extracted in three ways;second,an improved NMFmethod achieves dimension reduction and ensures the uniqueness results in dimensionality reduc-tion,avoiding the direct selection of the characteristics;finally,the SVM methods is used forcharacteristics identification and feature classification.Results show that the INMF method ex-tracts the weak signal characteristics effectively and improves the feature extraction accuracy ratein electromagnetic acoustic system.