Mitigating Cross-Database Differences for Learning Unified HRTF Representation

This is the official implementation of the WASPAA paper "Mitigating Cross-Database Differences for Learning Unified HRTF Representation."

This is also a follow-up work to our ICASSP 2023 paper "HRTF Field."

How to run our code

norm way 4 is the proposed method,

norm way 5 is mix_pos_mix_ear,

norm way 6 is mix_ear_individual_pos,

norm way 2 is equator.

Please modify the output folder (-o dir) and include the norm way info (--norm_way norm_way).

Experiments

experiment 1

python3 hrtf_normalization/train.py -o hrtf_normalization/exp1 -n listen crossmod sadie bili -t ari --norm_way 2

experiment 2

python3 hrtf_normalization/train.py -o hrtf_normalization/exp2 -n ari crossmod hutubs cipic -t riec --norm_way 2

experiment 3

python3 hrtf_normalization/train.py -o hrtf_normalization/exp3 -n crossmod sadie bili -t listen --norm_way 2

experiment 4

python3 hrtf_normalization/train.py -o hrtf_normalization/exp4 -n hutubs 3d3a bili listen crossmod cipic ita sadie ari

experiment 5

python3 hrtf_normalization/train.py -o hrtf_normalization/exp5 -n listen 3d3a bili crossmod cipic ita sadie ari riec -t hutubs --norm_way 2

SVM classification of different databases

To reproduce the result in Section 3 of our paper, simply run the following:

python3 database_classification.py

Citation

@INPROCEEDINGS{wen2023mitigate,
  author={Wen, Yutong and Zhang, You and Duan, Zhiyao},
  booktitle={Proc. IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)}, 
  title={Mitigating Cross-Database Differences for Learning Unified {HRTF} Representation}, 
  year={2023},
  volume={},
  number={},
  pages={1-5},
  doi={10.1109/WASPAA58266.2023.10248178}}

@INPROCEEDINGS{zhang2023hrtf,
  author={Zhang, You and Wang, Yuxiang and Duan, Zhiyao},
  booktitle={Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={{HRTF} Field: Unifying Measured {HRTF} Magnitude Representation with Neural Fields}, 
  year={2023},
  volume={},
  number={},
  pages={1-5},
  doi={10.1109/ICASSP49357.2023.10095801}}