Study On Customization Method Of Head Related Transfer Function

Human have the ability of perceiving sound source location which provides important additional information to vision.People utilize head related transfer function to filter sound source to synthesize sound image,so that the direction of sound source could be perceived.However,due to the differences between individual physical features,HRTF vary significantly among people,and are not suitable for everyone.In order to make the direction of sound image perceived more precisely, the work of customization needs to be done.This thesis proposed two HRTF customization methods. The methods are simply described as follows,(1 Use spectrum peaks and notches as localization cues in vertical directions,and use interaural time difference and interaural level difference as localization cues in horizontal directions.Genetic algorithm is adopted to search the customized parameters of spectral peaks and notches.Here subject’s response is used to decide the fitness value of candidates in current population.Interaural localization cues are adjusted through tuning the head radius which decide the interaural cues by two approximation expressions.The filter used is very simple, which is good for the real-time realization of 3D audio system, and the cost of customization is low.(2) Use principle component analysis to reduce the dimension of the HRTF magnitude response. Six principle components are selected to approximate the original magnitude response.Minimum phase reconstruction is applied to the customized magnitude response. As the method one genetic algorithm is adopted to search the principle component parameter space and the interaural time difference is tuned by hand to compose a complete HRTF. This method is faster than method one for the reason that less parameters need be tuned.