摘要
战场上车辆声音信号的构成非常复杂,采用单一的特征很难全面反映其特点,提取多种特征来构成特征向量就显得非常重要。应用改进的横虚警率检测(CFAR)算法对车辆声信号进行了分离,得到了数据的有效部分;提取了谐波集,Mel倒谱系数(MFCC)和小波能量3种特征,并应用主成分分析(PCA)方法对特征集进行了降维融合处理。实验结果表明:3种特征融合后的分类性能要好于单一特征,目标的识别率能够达到90%以上。
Vehicle acoustic signals in battlefield, which consist of many different components are very complex. Because a single acoustic feature can hardly reflect full characteristics of the vehicle, muhiple features should be extracted to form charaeteristic vector. Vehicle acoustic signals are separated from all the acquired signals by using modified CFAR algorithm. The three features are extracted,including harmonic set, MFCC and wavelet energy. But the resulting feature vector is too large, so PCA method is applied to reduce the dimension of feature vector. The experiment results show that the combined three features are better than the single feature in classification Derformance and the identification rate can reach above 90 %.
出处
《传感器与微系统》
CSCD
北大核心
2010年第7期30-32,共3页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(60575027)
关键词
特征提取
横虚警检测
特征降维
目标识别
feature extraction
constant false alarm rate（CFAR）
dimension reduction
target identification
