一种基于船舶辐射噪声信号改进Mel倒谱系数的目标识别方法
发布时间：2018-09-19 09:40
【摘要】：基于船舶辐射噪声信号Mel频率倒谱系数(MFCC)的目标类型识别是目前研究的一个热点。现有方法虽然在无噪声环境下具有较好的识别效果,但是在信噪比较低时其识别效果较差。基于此,文章提出了一种改进的提取MFCC特征参数的船舶目标识别方法,该方法在船舶辐射噪声信号的预处理阶段采用多正弦窗来代替传统使用的Hamming窗进行多窗频谱估计,经过计算得到改进的MFCC参数。试验结果表明,相比传统方法提取的MFCC参数,使用该方法提取的MFCC参数分别在不同信噪比的高斯白噪声干扰下,在BP神经网络分类器中的识别率更高,抗噪声的鲁棒性和稳定性更好。
[Abstract]:Target type recognition based on Mel frequency cepstrum coefficient (MFCC) of ship radiated noise signal is a hot topic. Although the existing methods have better recognition effect in noise-free environment, the recognition effect is poor when the signal-to-noise ratio (SNR) is low. Based on this, an improved ship target recognition method based on extracting MFCC characteristic parameters is proposed. In the preprocessing stage of ship radiated noise signal, multi-sinusoidal window is used instead of the traditional Hamming window to estimate the multi-window spectrum. The improved MFCC parameters are calculated. The experimental results show that, compared with the MFCC parameters extracted by the traditional method, the MFCC parameters extracted by this method have a higher recognition rate in the BP neural network classifier under the interference of Gao Si white noise with different signal-to-noise ratio (SNR), respectively. The robustness and stability of anti-noise are better.
【作者单位】： 江苏科技大学电子信息学院;
【基金】：国家自然基金项目(11574120) NSFC通用技术基础研究联合基金(U1636117)
【分类号】：U661.44;TN912.34
[Abstract]:Target type recognition based on Mel frequency cepstrum coefficient (MFCC) of ship radiated noise signal is a hot topic. Although the existing methods have better recognition effect in noise-free environment, the recognition effect is poor when the signal-to-noise ratio (SNR) is low. Based on this, an improved ship target recognition method based on extracting MFCC characteristic parameters is proposed. In the preprocessing stage of ship radiated noise signal, multi-sinusoidal window is used instead of the traditional Hamming window to estimate the multi-window spectrum. The improved MFCC parameters are calculated. The experimental results show that, compared with the MFCC parameters extracted by the traditional method, the MFCC parameters extracted by this method have a higher recognition rate in the BP neural network classifier under the interference of Gao Si white noise with different signal-to-noise ratio (SNR), respectively. The robustness and stability of anti-noise are better.
【作者单位】： 江苏科技大学电子信息学院;
【基金】：国家自然基金项目(11574120) NSFC通用技术基础研究联合基金(U1636117)
【分类号】：U661.44;TN912.34
【参考文献】
中国期刊全文数据库 前3条
1 陈迪;龚卫国;李波;;噪声鲁棒性说话人识别语音高频加权MFCC提取[J];仪器仪表学报;2008年03期
2 吴红卫;吴镇扬;赵力;;基于多窗谱的心理声学语音增强[J];声学学报;2007年03期
3 陆振波,章新华,朱进;基于MFCC的舰船辐射噪声特征提取[J];舰船科学技术;2004年02期
【共引文献】
中国期刊全文数据库 前10条
1 宣传忠;武佩;张丽娜;马彦华;张永安;邬娟;;羊咳嗽声的特征参数提取与识别方法[J];农业机械学报;2016年03期
2 曾以成;陈雨莺;毛燕湖;谢小娟;;基于经验模态分解结合傅氏变换与Wigner分布的Mel频率倒谱系数提取[J];湘潭大学自然科学学报;2015年02期
3 张贺;沈天飞;滕秋霞;;小词汇量孤立词语音识别系统多种特征组合参数的选择方法研究[J];电子测量技术;2015年03期
4 李响;谭南林;李国正;郭然;;一种应用语音多特征检测驾驶疲劳的方法[J];仪器仪表学报;2013年10期
5 陈冬;李钢虎;赵亚楠;;基于MVDR的MFCC方法在水下目标识别中的应用[J];声学与电子工程;2013年03期
6 王s，
本文编号：2249756
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2249756.html
