Research Of Technologies Of Sound Analysis/Synthesis And Simulation For Flight Simulator

With the rapid development of world civil aviation industry, espeacially inChina, demand for pilots exceeds supply largely. The flight simulator has become anecessary device to train pilots for airline companies and flight schools. Throughsimulating the feeling of vison, hearing, motion and force during flight, the flightsimulator can build a realistic flight environment on the ground. Sound simulationsystem is a key component of flight simulator and it can supply realistic andimmersive hearing for pilots. Sound simulation system can help pilots to judge theflight states through hearing for realizing real-time simulation of operating anaircraft. With the continuous development of computer science and technology,software development technology of sound simulation system has improvednoticeably and gained very high modularity, integration and portality. Comparing tosoftware which is easily transplanted and integrated, sound source extraction frommixed raw recording is necessary for each aircraft type. Therefore, sound sourceextraction wastes time and energy, and it is very expensive. Based on thebackground described above, our goal is to develop a sound simulation system withlow cost and high performance. We research the technology of sound analysis andsynthesis for sound source extraction and develop a sound simulation system forflight simulator. The final sound simulation system is integrated in the flightsimulator.Firstly, a sound source extraction method using short-time Fourier analysis andsynthesis is presented and we use this method to extract most of normal soundsources for sound simulation. We extract the sound of lowering and retractinglanding gear, and the sound of engine start successfully from the original recordingsin sound experiments. Through the comtrastive analysis, this method is feasible.Secondly, a noise source extraction method using linear prediction analysis andsynthesis is presented for many kinds of noise during flight. A linear predicitonall-pole model with white noise excitation source of zero mean and unit variance isbuilt. We use the autocorrelation method to deduce linear equations. After the linearequations are represented in matrix form, the Toeplitz autocorrelation matrix isgenerated. We use the method of Levinson-Durbin recursion to solve the linearequations. Comparing with traditional Gauss elimination, this method can not onlysave computation cost and storage, but also supply reflection coefficients which arethe intermediate quantities for lattice form realization of linear prediction synthesis.Through the sound experiment and comtrastive analysis, the method is better thanshort-time Fourier technology for noise.In sound simulation, some sounds of complex components, such as enginesound, may be necessary to further decompose. Therefore, a sound source extraction method based on spectral modeling synthesis technology is presented. Serra andSmith presented a spectral modeling synthesis technology which can realize a realseparation of sound sources. This method can decompose a sound into thedeterministic signal and noise signal. After that, we can extract the two signals. Thistechnology has good performance, but it has complicated theory and manyrealization steps need complex signal processing algorithms. Therefore, thistechnology is simplified according to a fact that aircraft engine sounds are usually acombination of harmonic sound plus random noise. The peak continuation algorithmis simplified to sine spectral synthesis after fundamental frequency detection. Theadditive synthesis in time domain is modified to high efficient frequency domainrealization based on inverse FFT. Line-segement approximation is replaced as linearprediction approximation. In a sound experiment, the sound of engine start isdecomposed into harmonic sound and noise from the original recording by usingsimplified SS spectral modeling synthesis technology, and the two sound sources areextracted. Through the comtrastive analysis, this method is feasible.A component scheme for sound simulation system is proposed. In thiscomponent, all data interactions and software realization are in main simulationcomputer, and sound computer is just a3D sound player. For this component design,an object-oriented program development method is used to realize software. Soundreal-time synthesis technology is used to solve the simulation of complexcomponents when the sounds are played. Audio API is used to solve soundspatialization. At last, the sound simulation system is connected seemlessly withother systems of flight simulator, and the objective of low cost and highperformance is met.Software realization method for developing sound simulation system in thisdissertation is easily updated, modified and transplanted. At the same time, the threemethods of sound source extraction which are presented here can be applied to manysound simulation system devlopments, not only for flight simulator, but also formany other simulators, such as cars, trains, tanks, armored cars.