HRTF and Neural Network Based Prediction and Simulation Method for Indoor Sports Acoustic

To obtain personally the hearing feeling on the virtual sound field in an interactive computer environment, this paper presents a HRTF and neural network based prediction and simulation method for indoor sports acoustic. The method obtains room impulse response of random position through frequency domain interpolation, realizes motion predicting mechanism utilizing the fore-and-aft relativity of the user´s motion position, and represents the motion of object by hierarchical data structure of quad tree. The mutual combination of the gradually coming in-out with head related transfer function HRTF and with neural network was accordingly turned. Multi-layer feed-forward neural network as the means to access any HRTF data in the indoors sound-field based on dynamic mechanisms was used. Protection of moving sound sources and room acoustic simulation based on dynamic mechanisms were determined the simulation of still sound source. Simulation results show that the method provides more reliable, the smoothly transition of the sequence of sound signals and the method of indoors acoustic simulation based on dynamic mechanism is feasible.