How to make closed-back headphones transparent for a vocalist's own direct sound

Park M, Ontakhrai S, Kittimathaveenan K, Alfredsson J, Ternström S. How to make closed-back headphones transparent for a vocalist's own direct sound. In: Express Paper. Audio Engineering Society; 2025. Paper 371. Available from: https://aes.org/publications/elibrary-page/?id=23045

@inproceedings{Park2025_23045,
  author = {Park, Munhum and Ontakhrai, Sorratorn and Kittimathaveenan, Kajornsak and Alfredsson, Jakob and Ternström, Sten},
  title = {{How to make closed-back headphones transparent for a vocalist's own direct sound}},
  note = {Paper 371},
  year = {2025},
  month = oct,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=23045}
}

TY  - CPAPER
TI  - How to make closed-back headphones transparent for a vocalist's own direct sound
AU  - Park, Munhum
AU  - Ontakhrai, Sorratorn
AU  - Kittimathaveenan, Kajornsak
AU  - Alfredsson, Jakob
AU  - Ternström, Sten
M1  - Paper 371
PY  - 2025
DA  - 2025/10/14
UR  - https://aes.org/publications/elibrary-page/?id=23045
PB  - Audio Engineering Society
LA  - en
AB  - In stage acoustics research, it is common to use virtual acoustic environments over headphones to simulate various room conditions for musicians. When experiments are conducted in suboptimal physical environments (e.g., without an anechoic chamber), it is often challenging to reduce the inherent reverberation of the test room while ensuring that musicians can hear their own direct sound through the headphones as if the headphones were transparent. In the present study, two methods were developed and tested using an acoustic dummy head, with the aim of faithfully replicating the direct sound of a solo singer over a pair of closed-back headphones. The results showed that the first method - creating and applying an exact finite-impulse-response (FIR) filter - may lead to undesirable effects, primarily due to the inherent delay of the playback system. The second method, which utilized a multiband equalizer, proved more effective when evaluated with stationary broadband noise. For non-stationary, real-world sounds such as singing and speaking voices, the comparison between the reference sound and the signal processed by the multiband equalizer remained reasonably accurate, with differences typically less than ~1 dB across much of the frequency range. Future work may further refine and evaluate the proposed methods through listening tests.
ER  -