摘要
传统的水声信号分类都是直接使用原信号进行处理的,特征提取耗时长,数据量大,针对这两个缺点,提出了一种压缩感知理论中基于稀疏表示的水声信号特征提取方法;该方法利用了水声信号在DCT变换域的稀疏特性,将信号的稀疏表示作为目标特征,并采用SVM分类算法进行分类识别。仿真结果表明,该方法不仅减少了特征向量的计算时间,还提高了目标分类识别率,还降低了水声信号的传输数据量,压缩率可达96%,在实际工程应用中具有较高的实用价值。
Traditional classification of underwater acoustic signal processing has the problems of time consuming feature extraction and large amount of data. A method based on compressed sensing theory using sparse representation of the acoustic signal feature extraction was introduced. This method utilized the acoustic signal sparse representation in the DCT transforming domain, and regarded the sparse representation of signals as the target feature, and then used SVM classification algorithm for target identification. Simulation results showed that, when using sparse representation as feature vectors, not only greatly reduced the time consuming features of computing, but also lifted effectively the underwater target recognition rate. At the same time, the amount of underwater acoustic data transmitting could be greatly reduced, and the compression rates were up to 96~.
出处
《探测与控制学报》
CSCD
北大核心
2014年第4期67-70,77,共5页
Journal of Detection & Control
关键词
压缩感知
稀疏表示
水声信号
特征提取
compressed sensing
sparse representation
underwater acoustic signal
feature extraction
