论文部分内容阅读
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。
The target type identification based on the Mel Frequency Cepstral Coefficients (MFCC) of ship radiated noise signal is a hot spot in the present research. Although the existing method has a good recognition effect in a no-noise environment, its recognition effect is poor when the signal-to-noise ratio is low. Based on this, an improved method of vessel target recognition based on MFCC feature parameters is proposed. The method uses multi-sine window instead of the Hamming window used in the pretreatment phase of ship radiated noise signal for multi-window spectrum estimation. Improved MFCC parameters are obtained. Experimental results show that MFCC parameters extracted by this method have higher recognition rate in BP neural network classifier than Gaussian white noise with different SNR, Better stick and stability.