摘要
根据人耳听觉特性,研究了心理声学中人耳识别目标的重要特征参数在目标识别中的应用。针对无线电噪声和舰船辐射噪声,利用Zw icker理论提取心理声学参数中的特性响度和特性尖锐度作为识别特征,通过神经网络分类器分别对这两组噪声各三类进行分类识别研究。实验表明特性响度和特性尖锐度主要反映了目标的振幅特性,可以正确识别目标并具有较高的识别率,是有效的识别特征。由于特性响度和特性尖锐度反映目标的特性相同,利用遗传算法仅对特性响度特征进行优化选择,挑选出特性响度中的分类关键量,降低识别空间的维数,提高识别率。
According to the characteristics of auditory system, application of the auditory psychoaeoustic parameters in target recognition is investigated. Firstly, specific loudness and specific sharpness as recognition features are calculated based on Zwicker＇ s theory. Secondly, the psychoacoustie parameters are used to classify the acoustic targets by using three - layer back - propagation neural network classifier. The simulations show that specific loudness and specific sharpness are effective features which reflect the amplitude characteristic of acoustic targets. Finally, because of the above features reflecting the same characteristic, the key features of specific loudness selected by genetic algorithm can reduce largely the recognition features＇ dimension and enhance the recognition rate.
出处
《计算机仿真》
CSCD
2008年第11期21-24,共4页
Computer Simulation
关键词
心理声学
特性响度
目标识别
Psychoacoustic
Specific loudness
Target recognition
