3D Sound Rendering Based On Individualized Head-Related Transfer Function

In daily life,people can decide which direction does a sound come from when hearing it.This is because human ear can sense the direction of sounds and then give a signal to brain.However,traditonal sound systems don't contain the spatial information of sounds.A 3-D sound system can reproduce the spatial information of sounds,and make the sounds more real.Binaural spatial system can reproduce spatial information of a sound source accurately by convolving it with Head-Related Transfer Function(HRTF)of left/right ear.HRTF can be seen as a filter between sound source and ear canal.HRTF is related to the size and shape of huaman body part like head,torso,shoulders,and pinnae,therefore,HRTF varies among different person.HRTF is also related to directions,HRTF of different dirctions are different.Using non-individualized HRTFs will cause some problems such as front-back confusion,up-down confusion,angular deviation etc,and affects the quality of the reconstructed spatial sound.Besides,common HRTF data has low spatial resolution,and needs a lot of memory to store.Moreover,the reconstructed 3D sound has inside-the-head-localizaiton(IHL)problem.This thesis conducts a series of research to solve these problems.(1)A novel HRTF personalization method based on anthropometric measurements is proposed.Based on the fact that HRTF is related to anthropometric parameters,several representive parameters are chosen,and then the best match of these parameters in database is found.The HRTFs of the best match subject are the final estimated individual HRTFs.(2)HRTF database consumes lots of memory.To reduce the memory size,a HRTF compression method is proposed.This method uses common factor decomposition(CFD)to deconvolve the HRTF,and then applies bamanced model truncate(BMT)to transfer the FIR filter into a low order IIR filter,which needs less memory.Experimental results show that this method can reduce memory size with little quality loss.(3)It is not practical to measure HRTF of every direction.Therefore,to construct HRTFs of new dirctions from known directions,a HRTF interpolation method is also proposed.(4)A artificial reverberator is added to the 3-D sound system to improve the annoying inside-the-head-localizaiton problem.The artificial reverberator has better performance and less computational complexity than others.Experimental results show that the "IHL" problem is less obvious than before.