Accelerating Audio Research with Robotic Dummy Heads

TL;DR

The paper addresses the bottleneck of slow, non-repeatable data collection in spatial audio research by introducing a robotic acoustic dummy head that moves quietly while recording binaural audio. It combines a 3D-printed head with a quiet, direct-drive motion platform to deliver realistic HRTFs and repeatable motion, validated against a KEMAR and demonstrated in moving-talker beamforming experiments using MVDR+CW. Key contributions include open-source design files, validated HRTF realism, and evidence of repeatable dynamic recordings suitable for objective evaluation of motion-aware algorithms, with potential to enable large-scale labeled spatial audio datasets. This tool promises to accelerate audio research across localization, tracking, source separation, and cocktail-party problem domains by enabling scalable, motion-enabled data collection without prohibitive costs or noise artifacts.

Abstract

This work introduces a robotic dummy head that fuses the acoustic realism of conventional audiological mannequins with the mobility of robots. The proposed device is capable of moving, talking, and listening as people do, and can be used to automate spatially-stationary audio experiments, thus accelerating the pace of audio research. Critically, the device may also be used as a moving sound source in dynamic experiments, due to its quiet motor. This feature differentiates our work from previous robotic acoustic research platforms. Validation that the robot enables high quality audio data collection is provided through various experiments and acoustic measurements. These experiments also demonstrate how the robot might be used to study adaptive binaural beamforming. Design files are provided as open-source to stimulate novel audio research.

Figures (11)

• Figure 1: Mannequins such as (a) and (b) are immobile, whereas the 3D printed dummy head is lightweight and compatible with a quiet motorized platform.
• Figure 2: HRTF magnitude in dB relative to 0$^\circ$ azimuth for a source 1.0 m away. The HRTFs of the KEMAR and printed head are alike at speech-relevant frequencies, indicating good acoustic realism. Note that HRTFs vary significantly across real human subjects.
• Figure 3: ILDs measured at 90$^\circ$ azimuth for the KEMAR, printed head, and retail mannequin. Comparison of the ILDs indicates that the printed head is acoustically realistic, based on its similarity to a KEMAR.
• Figure 4: Radiation patterns of the printed head (solid) and KEMAR (dashed), in dB relative to 0 $^\circ$ azimuth, for three octave bands.
• Figure 5: Servomotor noise is broadband and has significantly higher intensity than the stepper motor when both driven at 1 rev/s. The author's speech, at a conversational level and recorded from 1.3 m, is included for reference.