PaperlayerView exact revision on arXiv

Source-linked AI summary

A Device to Control and Manipulate Occlusion Effects for Own Voice Perception Studies

Rouben Rehman, Simon Kersten, Aron Schliep, Janina Fels

arXiv:2609.12845v1eess.AS

TL;DR

Perceptual research on the occlusion effect needs precise, independent control of insertion loss and occlusion gain. The authors use filtered headphone playback to emulate target conditions, and report that the system sufficiently matches target transfer functions.

  • Problem

    Perceptual work has given limited attention to principles linking occlusion-effect-related own-voice changes to communication behavior, while precise studies require independent control of insertion loss and occlusion gain.

    Source: §1 · pp. 1–2

  • Method

    A custom headphone system reproduces an appropriately filtered version of the user’s speech in real time, with filters derived by regularised least-squares optimisation.

    Source: §1 · p. 3, §5 · p. 15

  • Results

    The system sufficiently matched target transfer functions, demonstrating independent emulation and control of insertion-loss and occlusion-gain conditions.

    Source: §4.4 · p. 14, §5 · p. 15

  • Takeaways & Limitations

    The approach lays groundwork for formal listening tests investigating perceptual principles underlying the occlusion effect.

    Source: §4.4 · p. 14

    Residual inherent occlusion gain for low voices may affect perceptual studies; the authors recommend screening participants and potentially excluding voices below 110 Hz, depending on study design.

    Source: §4.4 · p. 14

Abstract

from arXiv · show

The occlusion effect (OE) refers to changes of the eardrum sound pressure through ear canal occlusion. It consists of two phenomena: an insertion loss (IL) attenuating air-conducted sounds, and an occlusion gain (OG) amplifying bone-conduction. Perceptual research on this is hindered by high variability of the OE across individuals, complicating repeatable presentation of precise OE conditions. Consequently, a method to control OE conditions reproducibly during perceptual experiments is needed. For such investigations, the system must enable separate control of IL and OG. We present an approach based on modified commercial earmuffs with integrated microphones. The design integrates dedicated impedance measurements and the derivation of digital filters, which are applied to the microphone signals to emulate arbitrary OE curves. The system is evaluated objectively through appropriate measurements. Results show that the headphones' inherent OG is guaranteed to be below 6dB above 115Hz, falling below 0dB above 155Hz. Emulation is shown to work accurately over the entire frequency range of interest. Limitations arise mainly due to a slight residual inherent OG for deep voices and the processing delay of the system. Future work will focus on the perceptual evaluation of the system and its application in perceptual studies.

1. Introduction

The occlusion effect changes eardrum sound pressure through air- and bone-conducted pathways, with its magnitude varying across devices and individuals. The paper presents a headphone approach to independently control insertion loss and occlusion gain, supporting reproducible perceptual studies.

1. Introduction

  • Occlusion changes eardrum sound pressure through air conduction and bone conduction, with device, insertion, seal, and ear anatomy affecting the resulting frequency ranges and amounts.Air-conducted sounds are attenuated by insertion loss, mainly above 1 kHz; occlusion gain amplifies bone-conducted sound below roughly 1 kHz.Source: §1 · pp. 1–2
  • Individual variability and the combined treatment of occlusion gain and insertion loss make controlled, reproducible perceptual studies difficult.The paper identifies this as limiting dose-response investigations and leaving changes in own-voice perception and communication behaviour understudied.Source: §1 · p. 2
  • The proposed system uses modified commercial earmuffs, external microphones, and internal transducers to reproduce the user’s voice in real time while physically attenuating air-conducted sound.Regularised least-squares filters derived for a desired target allow emulation of arbitrary occlusion-effect curves, with amplification and attenuation bounded by the transducer and earmuff, respectively.Source: §1 · pp. 2–3
  • Earlier earmuff setups studied bone-conducted own-voice transmission but were large and heavy enough to require external fixation, limiting participant movement.The authors identify natural communication behaviour as the motivation for redesigning the setup.Source: §1 · p. 3
  • The approach aims to support future investigations of cause-and-effect relations between the occlusion effect and own-voice perception, speech production, and communication behaviour.The introduction identifies precise control of both occlusion gain and insertion loss as necessary for perceptual studies.Source: §1 · pp. 2–3

2. Methodology

The method uses modified commercial earmuffs to reduce air-conducted sound and inherent occlusion gain, then designs filters to emulate target occlusion effects. Its design addresses attenuation, impedance, reproducibility, comfort, and real-time use.

2.1. Design Requirements

  • The headphones should attenuate air-conduction sounds by at least 20 dB above 500 Hz and 15 dB from 100–500 Hz, enable controllable, repeatable occlusion-effect emulation, and limit latency to 10 ms.They should be comfortable over longer periods without restricting participants’ movements too severely.Source: §2.1 · p. 3

2.2. Overall System Design

  • The design adapts commercially available earmuffs with sealed, modular cups and custom extenders to support high attenuation and reproducible construction.The extender sections and mounting brackets were designed for 3D printing.Source: §2.2 · pp. 3–4

2.3. Maximising the Inherent Insertion Loss

  • The earmuff design uses concentric inner and outer extender shells, mineral wool, sealing, and cavity damping to limit sound transmission.The shells avoid rigid connections that could transmit vibration, and their construction was designed to hinder transmission through the structure.Source: §2.3 · pp. 4–5

2.4. Minimising the Inherent Occlusion Gain

  • Occlusion gain depends on load impedance rather than enclosed volume alone, so the design matched headphone impedance toward the open-ear case.The authors used a transmission-line model to compare geometry permutations and estimate an upper bound on inherent gain.Source: §2.4 · pp. 5–6

2.5. Transducer and Microphone Inclusion

The system integrates a transducer and microphones into the earmuff design to reproduce the user’s filtered speech. The microphone mounts support repeatable positioning and detachment.

2.5. Transducer and Microphone Inclusion

  • A 40 mm driver was fitted to the inner extender shell to reproduce the user’s filtered speech in real time.The driver was selected based on its use in prior scientific work, where it showed a satisfactorily flat frequency response when equalised.Source: §2.5 · p. 6
  • Two MEMS microphones, one per side, were mounted on detachable brackets designed for precise and repeatable positioning.The microphone positions are marked in Figure 4.Source: §2.5 · p. 6
  • The microphones can be detached without sacrificing positional accuracy, supporting repeatable setup between measurements.The bracket slides over the frontal screw posts of the outer extender shell.Source: §2.5 · p. 6

2.6. Occlusion Emulation Filter Derivation

The filter derivation expresses open-ear and occluded-ear eardrum pressures through transfer functions, then finds a filter that makes the headphone system match a target occlusion effect. Required air-conduction transfer functions were measured using an artificial head and repeated repositioning.

2.6.1. Mathematical Derivation

  • The target occlusion effect is expressed using open- and occluded-ear transfer functions relative to a reference point, incorporating both magnitude and phase.The derivation approximates the open-ear eardrum pressure using its air-conducted component, whose bone-conducted component has little influence on own-voice perception.Source: §2.6.1 · p. 7
  • The filter G is chosen so the headphone system transfer function matches the target, using regularised least-squares optimisation because a direct solution would yield an unstable IIR filter.The derivation assumes leakage is negligible and the headphones introduce no significant inherent occlusion gain.Source: §2.6.1 · pp. 7–8

2.6.2. Obtaining the Required Transfer Functions

  • An artificial head with a mouth simulator was used to measure the air-conduction transfer functions directly, with repeated measurements involving microphone or headphone repositioning.The same setup was used for live validation; microphone frequency response was checked across 100 Hz to 8 kHz.Source: §2.6.2 · pp. 8–9

2.7. Real-Time Capability

Live filtering used a desktop PC, audio interface, and Reaper with ReaVerb. The filters were precomputed as 256-tap impulse responses at 44.1 kHz with a buffer size of 32.

2.7. Real-Time Capability

  • Signal processing was performed on a modern desktop PC with a Yamaha Steinberg UR44C audio interface, aiming to keep system complexity low for reproduction elsewhere.Source: §2.7 · p. 9
  • Live filtering used Reaper’s built-in ReaVerb plugin, with Reaper’s audio-thread priority set to MMCSS / Time Critical.Source: §2.7 · p. 9
  • Filters were precomputed as 256-tap impulse responses at 44.1 kHz with a buffer size of 32, the lowest tap length reported to reliably match target occlusion effects.Source: §2.7 · p. 9

3. Technical Evaluation Procedure

The evaluation measured inherent IL and OG, tested emulation against theoretical OE targets, and measured real-time latency. The emulation tests used artificial-head measurements, while latency was compared with processing removed.

3.1. Inherent Insertion Loss

  • The inherent IL was measured on a GRAS 45CA fixture against an empty-fixture reference, with headphones removed and repositioned between measurements.Figure 7 presents the mean IL in bold and standard deviation in shading.Source: §3.1 · p. 9, Figure 7 · p. 11

3.2. Inherent Occlusion Gain

  • Open-ear radiation and headphone load impedances were measured at three tube positions, then compared to calculate maximum inherent OG using Equation 5.Source: §3.2 · pp. 9–10

3.3. OE Emulation

  • Ten minimum-phase targets paired OG levels of 10, 15, or 20 dB with IL levels of −10, −15, or −20 dB, plus a flat 0 dB target.Source: §3.3 · p. 10
  • Emulation filters were tested with live filtering on an artificial head and compared against ideal target transfer functions using three repeated measurements.Source: §3.3 · pp. 10–11

3.4. Real-Time Capability

  • Around 6.5 ms was the consistent round-trip latency, measured with real-time filtering active and then with processing removed.Source: §3.4 · p. 11

4. Results and Discussion

The system matched arbitrary combinations of insertion-loss and occlusion-gain targets, with increased high-frequency measurement variance but continued target tracking after smoothing. Its measured inherent occlusion gain and insertion loss met the design’s stated conditions, while low-frequency residual gain and processing delay remain perceptual caveats.

4.1. Inherent Insertion Loss

  • Above approximately 150 Hz, the headphones’ inherent IL exceeds −20 dB, supporting emulation of ILs down to −20 dB and lower depending on the target.The measured IL remains at or below −18 dB across the frequency range, preserving at least 10 dB separation between emulated and leakage components.Source: §4.1 · p. 11

4.2. Inherent Occlusion Gain

  • Measured load impedances differed negligibly between sides, and measurements and transmission-line simulations generally agreed, supporting the model’s use in headphone design.The measured standard-deviation regions never exceeded 2.26 dB; load impedance matched the open-ear radiation impedance within 3 dB above about 150 Hz and within 1.4 dB above 400 Hz.Source: §4.2 · pp. 11–12
  • The headphones’ inherent own-voice OG reaches an absolute maximum of 8.7 dB at 100 Hz; its upper bound reaches 0 dB around 155 Hz and is guaranteed below 0 dB thereafter.At low-voice frequencies, including the male-voice mean of 116 Hz, OG is guaranteed below 6 dB. Studies should screen and potentially exclude participants below 110 Hz, depending on design.Source: §4.2 · p. 12

4.3. OE Emulation

  • The system matched arbitrary IL–OG target combinations; despite rising variance above 2.5 kHz, 1/6-octave-smoothed responses still tracked targets.The authors conclude the approach can sufficiently match all investigated targets for the intended application and independently control IL and OG.Source: §4.3 · p. 13, Figure 10 · p. 14

4.4. Limitations and Suggestions for Future Work

  • The inherent-OG estimate assumes headphones do not alter bone-conduction transmission to the outer ear; the authors use this assumption to derive an upper bound.They argue it is suitable for headphones positioned over the ears, while noting it does not fully describe arbitrary occlusion devices.Source: §4.4 · p. 14
  • A perfect impedance match down to 100 Hz was not achieved while retaining a lightweight, self-contained design, leaving residual OG for low voices.Depending on study design, participants with voices below 110 Hz may need screening and potential exclusion.Source: §4.2 · p. 12, §4.4 · p. 14
  • Perceptual scope remains untested: the emulation uses a time-invariant model, and system delay may slightly alter own-voice timbre.The authors recommend formal listening tests, future integration of time-varying effects, and lower-latency hardware.Source: §4.4 · p. 14

5. Conclusion

The custom headphone system sufficiently matched target transfer functions, supporting independent control of insertion loss and occlusion gain. Residual inherent occlusion gain for deep voices and processing delay remain limitations.

5. Conclusion

  • The system sufficiently matched target transfer functions for independent emulation and control of insertion loss and occlusion gain.Evaluated targets included OGs of 10 dB, 15 dB, and 20 dB; ILs of −10 dB, −15 dB, and −20 dB; and a 0 dB open-ear target.Source: §5 · p. 15
  • The system reproduces an appropriately filtered version of the user’s speech in real time using custom headphones based on commercial earmuffs.Source: §5 · p. 15
  • A slight but perceptible inherent OG for deep voices and processing delay potentially altering own-voice timbre are the main limitations.Source: §5 · p. 15
  • Future work will evaluate the system perceptually, extend it to time-variant OE emulation, and implement the DSP chain on dedicated low-latency hardware.Source: §5 · p. 15

Author contribution statement

The contribution statement assigns conceptualization, investigation, and methodology to multiple authors, while other roles are distributed across the team.

Author contribution statement

  • Rehman contributed to conceptualization, data curation, formal analysis, investigation, methodology, software, visualization, and the original draft.Source: Author contribution statement · p. 15
  • Kersten contributed to conceptualization, investigation, methodology, and review and editing; Schliep contributed to investigation, methodology, visualization, and review and editing.Source: Author contribution statement · p. 15
  • Fels contributed to conceptualization, project administration, funding acquisition, supervision, and review and editing.Source: Author contribution statement · p. 15

Appendix A: Transmission Line Model

The appendix models the headphone extender and earmuff capsule acoustically, then selects extender geometry as a compromise between inherent OG and headphone size and weight.

Appendix A: Transmission Line Model

  • The transmission-line model represents the extender with connected lines and conical horns containing polyester filling, and the earmuff capsule with an acoustic-foam-filled line and rigid termination.Source: Appendix A · p. 17, Figure A.1 · p. 18
  • The middle extender section’s cross-sectional area and length were varied while its ends remained constant; Figure A.2 presents simulated load impedances and maximum inherent OG across geometries.Source: Appendix A · p. 17, Figure A.2 · p. 18
  • The selected geometry was 3 cm long with a 60 cm2 area, judged the best compromise between inherent OG and physical size and weight; the final middle section was 3 cm and 59.1 cm2.Source: Appendix A · p. 17
  • The final inner extender section measured 11.5 cm overall, with cross-sectional area varying from 30 cm2 to 59.1 cm2.Source: Appendix A · p. 17
  • The model used the Johnson-Champoux-Allard equivalent-fluid model for polyester filling and acoustic foam, with foam airflow resistivity adjusted to match initial impedance measurements.Source: Appendix A · p. 17, Table A.1 · p. 17, Table A.2 · p. 17

Appendix B: Least-Squares Optimisation

The appendix formulates filter derivation as a least-squares problem using convolution matrices, then regularizes the solution to address ill-conditioning.

Appendix B: Least-Squares Optimisation

  • The desired OE, filter, extender, and microphone impulse responses are represented as vectors, and the optimal filter is obtained by minimizing a cost function.Source: Appendix B · pp. 19–20
  • Convolution matrices for the extender and microphone responses, together with a zero-padded OE matrix, express the filter-design problem in matrix form.Source: Appendix B · p. 19
  • Zero-padding the OE matrix matches the dimensions of the microphone-extender matrix product in practice.Source: Appendix B · p. 19
  • The unconstrained optimal solution projects the target vector onto the column space of the design matrix, while regularization addresses the typically ill-conditioned matrix inversion.Source: Appendix B · p. 20
Loading 2609.12845v1…