听觉模型输出谱特征在声目标识别中的应用

Application of auditory spectrum-based features into acoustic target recognition

利用模拟人耳声信号处理过程的CcGC滤波器组模型,研究了听觉特征应用于声目标识别相比传统特征的优势。结果表明:当信号的信噪比下降时,听觉特征逐渐表现出更好的性能,体现了听觉系统优异的抗噪声能力。随后,本文从CcGC滤波器组模型所反映的听觉系统四个主要特性入手,通过仿真实验研究了耳蜗抑制噪声的机理,结果表明听觉系统的临界带划分和非线性压缩在耳蜗抑制噪声中起着关键的作用。

In the acoustic target recognition, the auditory feature＇s advantages compared with the traditional feature are studied by using the cascade compressive gammachrip （CcGC） filter model, which has been used to describe auditory system function. It is concluded from theoretical analysis and simulations that the auditory feature has better effects as the Signal-to-Noise Ratio （SNR） is reduced. At last, the mechanism of noise suppression in cochlea is analyzed by simulations, and it is shown that Equivalent Rectangular Bandwidth （ERB） expression in frequency domain and the compression of cochlea are essential for cochlea noise suppression.