AN INTEGRATED CNN-LSTM METHOD TO FAR-FIELD SOUND PREDICTION USING NEAR-FIELD MEASUREMENTS

The far field refers to the acoustic region far from the sound source, however, measuring the far-field sound requires a lot of measurement mics and may sometimes be impossible in practice. Specifically, sound reconstruction and prediction usually rely on physical-based methods to estimate the sound pressure in the target zone. In this study, we propose an artificial intelligent algorithm to evaluate far-field sound using the near-field measurements. Incorporating convolutional neural network (CNN) and long short-term memory (LSTM), the architecture is capable of predicting sound in both spatial and temporal domains. Experimental results show that the proposed method can accurately predict the far-field magnitude of wideband sound pressure in real time.