BUJEON AI-Ear · Spatial intelligence

Human hearing.
For robots.

HRTF-based biomimetic two-channel sensors and BiSELD.
Jointly estimating sound types and 3D locations for multiple sources.

Explore the space of sound

TWO CHANNELS. FULL HEMISPHERE.Left + Right · ON

01 Speech

Azimuth: 30°
Elevation: 32°
Distance: 3.2 m

02 Impact

Azimuth: -74°
Elevation: 48°
Distance: 5.8 m

03 Alarm

Azimuth: 137°
Elevation: 24°
Distance: 7.4 m

01 / 04Sound onset

01Dual input→02Spatial cues→03Type + 3D location

Conceptual visualization of moving sound sources and receiver modules. Not measured data or inference results.

BINAURAL SCENE UNDERSTANDING

Sound type.Event timing.Spatial coordinates.

From two-channel input to scene information.

BINAURAL CUES

One sound,
two different signals.

Small differences between the two channels carry spatial information.
We analyze timing, sound level, and frequency together to interpret the direction of sound.

01 Δt

Arrival-Time Differences

The same sound reaches the two channels at different times. We analyze the time difference between corresponding signals as a cue to estimate direction.

02 ΔL

Level Differences

Sound levels at the two channels vary with the direction of arrival. We use the relative level difference between channels as a spatial cue.

03 ƒ

Spectral Differences

The head and outer ears shape sound transmission across frequencies. We interpret these direction-dependent patterns as HRTF-based spatial cues.

A synthetic visualization illustrating differences between binaural signals. Not actual measurements or inference results.

BINAURAL INPUT / DUAL CHANNEL

Two-channel differences,
inputs to spatial estimation.

We compare the binaural signals received by Left and Right. Timing, level, and spectral differences provide spatial cues for BiSELD to jointly estimate the types and 3D locations of multiple events.

HRTF-based biomimetic sensingINPUT → ANALYSIS → INFERENCE

LeftLeft input

RightRight input

JOINT INFERENCEBiSELD

Event classification
Spatial localization

OUTPUTEvents and coordinates

Event type

Multiple sound event classification

3D location

Azimuth · Elevation · Distance

Binaural spatial cues

ΔtInteraural time differenceΔLInteraural level differenceƒSpectral characteristics

Conceptual illustration of binaural signal differences. Waveforms and processing flows are illustrative, not measured data or inference results.

SED / EVENT ONSET & OFFSET

When does it start,
and when does it end?

Sound event detection (SED) estimates event types and their onset and offset times. Marking event intervals on a timeline reveals durations and overlaps between multiple events.

SED / SOUND EVENT DETECTION12 s / SYNTHETIC SEQUENCE

START · ONSETDURATION · ACTIVEEND · OFFSET

Event classificationFootsteps · Conversation · Ringing

Time boundariesOnset time · Offset time

Overlapping intervalsSimultaneous sound events

A synthetic sequence illustrating onset and offset detection in SED. Not an actual recording or inference result.

BUJEON R&D / HRTF MEASUREMENT

Foundations of spatial hearing,
HRTF research infrastructure.

BUJEON measures direction-dependent head-related transfer functions (HRTFs) using a KEMAR dummy head in an anechoic chamber. We record changes in the signals reaching both ears to provide foundational data for spatial hearing research.

BUJEON ANECHOIC CHAMBER

BUJEON anechoic chamber · KEMAR-based HRTF measurement setup

01 / MEASURE

Binaural Transfer Characteristics

We measure changes in timing, sound level, and spectrum caused by the head and outer ears, recording signal transfer characteristics at both ears for each direction.

02 / MAP

Directional HRTF Measurement

We build a far-field HRTF database through measurements at 5° intervals over 360° of azimuth and elevations from −80° to +90°.

03 / VALIDATE

Measurement vs. Simulation

We study direction-dependent acoustic transfer characteristics by comparing measured HRTFs with numerical results from the boundary element method (BEM).

Frequency patterns from two channels, updated over timeIllustrative synthetic patterns · Not actual measurements

HRTFs describe direction-dependent sound transmission shaped by the head and ears. AI-Ear uses HRTF-based biomimetic two-channel sensors and BiSELD to jointly estimate the types and 3D locations of multiple sound sources.

HRTF → 2 CHANNELS → BiSELD

BUJEON AI-EARRESEARCH PERSPECTIVE

RESEARCH OUTLOOK / Future Direction

Audio × Vision

A path to deeper understanding.

Building on our spatial acoustics expertise,
we plan to integrate Vision AI.
Connecting different cues broadens our understanding of a scene.

01 / AUDIOTechnology Foundation

Spatial Acoustic Analysis

Using cues from two channels,
we interpret sound types and locations.

02 / VISIONPlanned Expansion

Connecting Visual Information

We plan to connect visual objects and motion
with audio information.

03 / MULTIMODALOur Aim

Integrated Scene Understanding

Our aim is multimodal AI
that interprets sound and scenes together.

From today's technology foundation toward future integration.

Discuss technical collaboration

BUILT ON AUDIO. OPEN TO VISION.BUJEON R&D

TECHNICAL INQUIRY

Spatial hearing.
Explore its applications.

Discuss applications for HRTF-based sensing and BiSELD,
system requirements, and research collaboration.

Contact the AI-Ear team ↗

Human hearing.For robots.