To generate effective 3D sound, HRTFs have been used to two channel speaker systems or headphone/earphone based VAD systems. We suggest two conceptual ideas about 3D sound reproduction based on HRTFs: HRTF customization and considering near-field HRTF. We focus on a feasibility of application of the two ideas to mobile devices. For human-mobile interaction, those will be able to provide users with proper sound cues. Moreover, we introduce sound source localization technique using TDOA of a pair of microphones. Using this technique in mobile device, the direction of user can be estimated with acceptable accuracy, and this can assist sound enhancement and visual system.