An application of the neural network in the near-field acoustic holography

The near-field acoustic holography (NAH) is an effective and useful noise measurement method. There are many methods for reconstruction of the sound source, but the measurement circumstance is limited and the random noise is unavoidable. In this paper, a new holography reconstruction method based on Neural Network (NN) is proposed, which can identify the sound source and restrain the random noise better.Neural Network (NN) is widely used in many fields recently. It has many main characters, such as the ability of learning and nonlinear approximation capability. In order to improve the measurement accuracy, a suitable net structure and learning algorithm of NN need to be found. The sound pressure distribution is measured as input and K-L transformation is applied to extract the input.Output is the distribution of the noise sound. Experiment results indicate that the method can identify the noise source effectively. Under the condition with random noise, it is also effective and accurate. In conclusion, the reconstruction method based on NN is more accurate and more applicable.