全息数据外推与插值技术的极限学习机方法:全息数据外推与插值技术的极限学习机方法
Hologram data extrapolation and interpolation method based on the extreme learning machine
下载全文
作 者:孙超[1,2] 何元安[3] 商德江[2] 刘月婵[2]
SUN Chao, HE Yuanan, SHANG Dejiang, LIU Yuechan ( 1.College of Power and Energy Engineering, Harbin Engineering University, Harbin 150001, China;2.College of Underwater Acoustic Engineering, Harbin Engineering University, Harbin 150001, China; 3. Shipping System Engineering Research Institution, Beijing 100036, China)
机构地区:[1]哈尔滨工程大学动力与能源工程学院,黑龙江哈尔滨150001 [2]哈尔滨工程大学水声工程学院,黑龙江哈尔滨150001 [3]船舶系统工程研究院,北京100036
出 处:《哈尔滨工程大学学报》 EI CSCD 2014年第35卷第5期 544-551页,共8页
Journal of Harbin Engineering University
基 金:国家自然科学基金资助项目(11274080).
摘 要:针对近场声全息技术中有限测量孔径问题,提出一种基于极限学习机的全息声压外推与插值方法。该方法将初始全息面上的测量数据作为网络的输入样本进行训练学习,将未知的全息面声场当做目标值进行预测,实现全息数据的外推与插值,并对影响外推与插值结果的参数进行分析。利用外推或插值后的全息数据进行重建,将重建结果进行比较。结果表明:该方法相对于传统方法获得了更高的重建精度,解决了测量孔径有限造成的较大重建误差,在不增加测点情况下提高了重建图像的空间分辨率。同时,该插值技术也可用于非自由声场环境中,仿真结果验证了该方法的稳健性和优越性。
Due to the finite measurement aperture in the near-field acoustic holography(NAH)technology,a meth-od for hologram data extrapolation and interpolation based on the neural network extreme learning machine was pro-posed. The network was constructed by treating the measurement data of the initial holographic plane as the training sample of the network and the acoustic field of the unknown holographic plane as the target value of the forecast, thus realizing extrapolation and interpolation of the holographic data,and then the parameters were analyzed that in-fluence the results of extrapolation and interpolastion.Comparisons of the reconstructions which using the extrapola-tion or interpolation holographic data,show that the reconstruction results obtained by this method achieve better sound field reconstruction accuracy as compared with the traditional method. The sound field reconstruction accura-cy can be effectively improved by the data extrapolation from small aperture measurement data. The accuracy and the resolution of the reconstruction with a few sensors can also be effectively improved by the data interpolation. This interpolation technology can also be used in a non-free acoustic field environment. The robustness and superiority are verified by the simulation results.
关键词:极限学习机 数据外推 数据插值 近场声全息 重建误差 测量孔径 空间分辨率
extreme learning machine ;data extrapolation ;data interpolation ;near-field holography ;reconstruction error ;measurement aperture ;field reconstruction accuracy ;