摘要
针对声源识别中观测模型线性化误差,信号特征参数提取依赖于经验分析阈值判断而造成信息丢失的问题,从信号观测模型、预处理、特征提取与分类识别、半实物仿真试验等方面,提出了一种新的四元阵列融合声源识别方法。首先在系统坐标系下建立了四元阵列有色噪声环境下的观测模型;其次基于经验模态分解(empirical mode decomposition,EMD)理论,给出了四元阵列EMD融合算法,有效抑制了高频信号的干扰;再次基于梅尔频率倒谱系数-动态时间规整(Mel-frequency cepstrum coefficient-dynamic time warping,MFCC-DTW)方法,设计了阵列信号特征提取与分类识别算法;最后通过半实物仿真试验,并与相关研究基础对比,分别验证了提出的EMD融合算法及阵列信号特征提取与分类识别算法的有效性。
In the sound source recognition,the observation model may have the linearization error and extraction signal characteristic parameters depended on the threshold value of empirical analysis may cause information loss.To deal with those problems,an innovative four-sensor acoustic array fusion identification algorithm was proposed from the aspects of signal observation model,preprocessing,feature extraction and classification recognition,and hardware-in-the-loop simulation experiments.Firstly,the observation model of the four-sensor acoustic array was established in the system coordinate system.Secondly,an empirical mode decomposition(EMD)fusion algorithm was given based on the EMD theory,which effectively suppressed the interference of high frequency signals.And then the algorithm of array signal feature extraction and classification recognition was designed based on the Mel-frequency cepstrum coefficient-dynamic time warping(MFCC-DTW)method.Finally,the effectiveness of the proposed EMD fusion algorithm and array signal feature extraction and classification recognition algorithms were verified through semi-physical simulation experiments and comparison with related research foundations.
作者
刘亚雷
顾晓辉
甘宁
LIU Ya-lei;GU Xiao-hui;GAN Ning(Department of Warship Command,China Maritime Police Academy of CAPF,Ningbo 315801,China;ZNDY of Ministerial Key Laboratory,Nanjing University of Science and Technology,Nanjing 210094,China)
出处
《科学技术与工程》
北大核心
2020年第28期11620-11625,共6页
Science Technology and Engineering
基金
国家自然科学基金青年基金(51605227)
公安部科技创新项目(2017JSYJC11)。
关键词
被动声识别
经验模式分解
梅尔倒谱参数
动态时间规整
passive acoustic identification
empirical mode decomposition
Mel-frequency cepstrum coefficient
dynamic time warping
