Reinforcement-Learning-Based Personalization of Head-Related Transfer Functions

Nambu I, Washizu M, Morioka S, Hasegawa Y, Sakuma W, Yano S, Hokari H, Wada Y. Reinforcement-Learning-Based Personalization of Head-Related Transfer Functions. J Audio Eng Soc. 2018;66(5):317-328. doi:10.17743/jaes.2018.0014

@article{Nambu2018_19563,
  author = {Nambu, Isao and Washizu, Manabu and Morioka, Shuhei and Hasegawa, Yuta and Sakuma, Wataru and Yano, Shohei and Hokari, Haruhide and Wada, Yasuhiro},
  title = {{Reinforcement-Learning-Based Personalization of Head-Related Transfer Functions}},
  journal = {Journal of the Audio Engineering Society},
  volume = {66},
  number = {5},
  pages = {317--328},
  year = {2018},
  month = may,
  publisher = {Audio Engineering Society},
  doi = {10.17743/jaes.2018.0014},
  url = {https://doi.org/10.17743/jaes.2018.0014}
}

TY  - JOUR
TI  - Reinforcement-Learning-Based Personalization of Head-Related Transfer Functions
AU  - Nambu, Isao
AU  - Washizu, Manabu
AU  - Morioka, Shuhei
AU  - Hasegawa, Yuta
AU  - Sakuma, Wataru
AU  - Yano, Shohei
AU  - Hokari, Haruhide
AU  - Wada, Yasuhiro
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 66
IS  - 5
SP  - 317
EP  - 328
PY  - 2018
DA  - 2018/05/06
DO  - 10.17743/jaes.2018.0014
UR  - https://doi.org/10.17743/jaes.2018.0014
PB  - Audio Engineering Society
LA  - en
AB  - In order to perceive spatial locations of virtual sounds using stereo headphones, individual head-related transfer functions (HRTFs) are required for each listener. However, accurate HRTF measurement is usually difficult. While previous studies have proposed methods of HRTF personalization without HRTF measurement, localization errors often remain and further modifications are challenging. This research proposes a method that uses reinforcement learning and listener evaluation to obtain an accurate individual HRTF without measurement. The authors conducted a proof-of-concept simulation with an experiment involving human subjects. In the simulation, it was confirmed that the proposed method could acquire individual HRTFs close to the measured dummy-head HRTF. A learning experiment in one direction used the proposed method without individual HRTFs. The results showed improved horizontal-plane localization for the learned HRTF as compared to the dummy-head HRTF. These experiments collectively demonstrate the possibility of the proposed reinforcement-learning-based personalization method for individual HRTFs that enables listeners to experience accurate virtual sound environments.
ER  -