数据融合技术在声全息测量中的应用

Application of data fusion to acoustic holography

提出了基于神经网络和证据理论的数据融合技术。首先,根据声全息的测量原理,建立了由传感器子网和融合子网组成的数据融合模型。接着,给出了基于神经网络的传感器子网结构,实现了从目标特征参数到目标类型的映射,得到初步的输出结果。然后,采用证据理论将目标信息融合起来,达到对目标的有效识别得到最后的识别结果。最后,给出了数据融合技术应用于声全息法识别声源的实例计算。实验结果表明:数据融合后的声源识别率为94.2%,比融合前提高了11.7%。该技术减小了由于信息量不足或存在较大偶然误差而带来的不利影响,使声源的识别结果更可靠。

A data fusion technology was proposed based on neural network and evidence theory, and a data fusion model consisting of the sensor subnet and the fusion subnet was established based on the measurement principle of acoustic holography. The sensor subnet on neural network was given to realize the map from the target characteristic parameters to the target type, so that the preliminary output was obtained. Then, the evidence theory and target information were synthesized to obtain the final identification result was obtained. Finally, an example was provided to illustrate the application of data fusion technology to the sound source identification of acoustic holography. Experimental results indicate that the identification rate of sound source is 94.2% after data fusion, It is increased by 11.7%. The application of data fusion technology can reduce the adverse effects caused by insufficient information or accidental errors and make the identification results of sound source more reliable.