DeepBSL: 3-D Personalized Deep Binaural Sound Localization on Earable Devices
Bibliographic record
Abstract
The prevalence of earable devices, such as earbuds and headphones, allows people to converse or listen to audio recordings on the move but often at the cost of reduced alertness of imminent health and safety threats in their surroundings. 3-D binaural sound localization (BSL), which aims to locate sound sources in space, plays a key role in improving one’s situation awareness. BSL on earable devices is inherently challenging due to the limited number of microphones available as well as subject- and location-dependent filtration effects of a person’s pinna, head and torso, described by head-related transfer functions (HRTFs). In this work, we develop DeepBSL, a deep neural network model to estimate azimuth and elevation angles of sound sources relative to a person’s head. For a new subject, an efficient procedure is developed to collect HRTFs at sparse locations using in-ear microphones and a mobile phone, which are then utilized to synthesize sounds of any type at arbitrary locations to train personalized DeepBSL models. Extensive evaluations using synthetic data from a public data set and through real-world experiments demonstrate that the personalized DeepBSL models are data-efficient and can achieve better-than-human performances in BSL while significantly outperforming a state-of-the-art model that can only predict azimuth angles of sound sources. Our best performing model has an average azimuth prediction error of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.9^{\circ } (4.1^{\circ })$ </tex-math></inline-formula> and elevation prediction error of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.4^{\circ } (2.9^{\circ })$ </tex-math></inline-formula> in an indoor (outdoor) environment.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".