论文部分内容阅读
为研究短波语音通信下的飞机识别,提出利用2种方式对目标声信号进行分析处理.为实现对语音进行抑制,分别利用全局经验模态分解(EEMD)和经验模态分解(EMD)将信号进行重构,然后根据重构后的目标信号进行Bark域频率感知的小波包分解(BWPD)和高阶累积量(HOC)分解,对目标声信号分别提取了听觉感知的特征和展现信号的物理特性的特征;分别利用EEMD和EMD分解对信号进行重构,然后选择Mel频率倒谱系数和高阶累积量对重构后的信号进行特征提取.对比实验表明:EEMD-BWPD-HOC方法能够抽取出有效的飞机舱内背景声音信号特征,实现语音抑制,并且以较高的识别率识别出4种飞机.
In order to study the recognition of aircraft in short-wave speech communication, two methods are proposed to analyze and process the target acoustic signal.In order to suppress the speech, the global empirical mode decomposition (EEMD) and empirical mode decomposition (EMD) Then the BWPD and HOC decomposition of the Bark domain are performed according to the reconstructed target signal, and then the characteristics of auditory perception and the characteristics of signal physics The EEMD and EMD decomposition are used to reconstruct the signal respectively, then the Mel frequency cepstrum coefficients and high-order cumulants are selected to extract the features of the reconstructed signal.Comparison experiments show that the EEMD-BWPD-HOC method can extract A valid background sound signal characteristics of aircraft cabin, to achieve voice suppression, and to identify a higher recognition rate of four aircraft.