水声目标的MFCC特征提取与分类识别
MFCC Feature Extraction and Classification of Underwater Acoustic Targets
葛轶洲 1姚泽 2张歆 2周青1
作者信息
- 1. 通信信息控制和安全技术重点实验室,浙江 嘉兴 314033;中国电子科技集团公司第三十六研究所,浙江 嘉兴 314033
- 2. 西北工业大学航海学院,陕西 西安 710072
- 折叠
摘要
水声目标识别技术在水下信息处理中起着非常重要的作用,从辐射噪声中提取水声目标的有效特征一直都是水声目标识别技术的难点所在.提出了一种利用水声目标辐射噪声的梅尔频率倒谱系数(Mel-Frequency Cepstral Coefficients,MF-CC)作为目标特征提取的方法.通过对辐射噪声信号进行梅尔频率滤波得到目标噪声信号的MFCC特征,它模拟了人耳对不同频率的声音具有不同感知能力的听觉非线性效应,因此具有良好的识别效果.通过对实际水声目标的辐射噪声进行测试实验,提取目标噪声信号的MFCC特征向量,并运用K近邻算法对其进行分类识别,实验结果显示MFCC特征提取与分类识别算法对水声目标的识别率达到85%以上.
Abstract
Underwater acoustic target recognition technology plays a very important role in underwater information processing.Extracting effective features of underwater acoustic targets from radiated noise has always been the diffi-culty of underwater acoustic target recognition technology.This paper proposes a method of using Mel-Frequency Cepstral Coefficients(MFCC)of underwater acoustic target radiated noise as target feature extraction.The MFCC characteristics of the target noise signal are obtained by filtering the radiated noise signal by Mel frequency.It simu-lates the auditory nonlinear effect that the human ear has different perceptual abilities of sounds of different frequen-cies,so it has a good recognition effect.Through the test experiment on the radiation noise of the actual underwater a-coustic target,the MFCC feature vector of the target noise signal is extracted,and the K-nearest neighbor algorithm is used to classify and recognize it.The experimental results show the recognition rate of the underwater acoustic target by the MFCC feature extraction and classification recognition algorithm Over 85%.
关键词
水声信息对抗/特征提取/梅尔倒谱系数/分类识别Key words
Underwater acoustic information confrontation/Feature extraction/MFCC/Classification and recogni-tion引用本文复制引用
出版年
2024
NETL
NSTL
万方数据