Personal Sound Zones Overview

• Personal Sound Zones are localized audio regions in shared environments that deliver interference-free, high-fidelity sound through precise active control and filtering.
• They utilize methods like pressure matching, acoustic contrast control, and neural network-based filter synthesis to achieve robust metrics such as inter-zone and inter-program isolation.
• Applications include targeted audio delivery in dynamic settings using parametric loudspeakers and adaptive algorithms that optimize spatial audio rendering in multipath and reverberant spaces.

Personal Sound Zones (PSZs) refer to localized regions within a shared acoustic environment where audio programs are rendered with high fidelity, isolation, and minimal cross-talk, enabling multiple listeners to experience personalized content without resorting to headphones. The PSZ concept encompasses active acoustic field control, distributed filtering, advanced nonlinear transducer designs, and neural-network-based filter synthesis, with quantifiable metrics such as inter-zone isolation (IZI), inter-program isolation (IPI), and acoustic contrast serving as key indicators of performance.

1. Theoretical Foundations and Zone Size Constraints

Physical constraints on PSZ formation derive from the spatial-temporal correlation properties of broadband diffuse sound fields. For a stationary diffuse field, the normalized pressure correlation $\rho(\Delta \mathbf r, \Delta t)$ between points is given by

$$\rho(\Delta \mathbf r, \Delta t) = \frac{1}{2\pi\,E[p^2]} \int_{-\infty}^{\infty} S(\omega)\, \mathrm{sinc}\left(\frac{\omega\,\Delta \mathbf r}{c}\right) e^{j\omega\,\Delta t} \, d\omega$$

where $S(\omega)$ is the spectral power distribution and $E[p^2]$ the total field variance (Rafaely, 2023). In classic pure-tone fields, the "zone of quiet" (defined as the region with at least 10 dB attenuation), has a radius $\Delta r_{10\,\mathrm{dB}} \approx 0.088 \,\lambda$, meaning that for a 1 kHz tone ($\lambda \approx 0.343$ m), the effective PSZ radius is ∼3 cm. Broadband noise extends the zone radius only modestly (up to 40%), with the strict limit set by the center frequency. This physical constraint applies regardless of control methodology, implying that any practical PSZ system must contend with a fundamental zone-size limitation.

2. Active Control Algorithms and Zone Shaping

Classical active control approaches optimize loudspeaker filter vectors $\mathbf{g}(f)$ to match target sound fields in the bright zone while suppressing energy in the dark zone. Standard formulations include:

• Pressure Matching (PM):

  $J(\mathbf{g}) = \|H\,\mathbf{g} - d\|^2$

  for system ATF $H$, desired pressures $d$, yielding filter updates via complex LMS or matrix inversion. Adaptive versions operate iteratively (Zhang et al., 2023).

• Acoustic Contrast Control (ACC):

  $$\max_{\mathbf{g}} \frac{\|H_b\,\mathbf{g}\|^2}{\|H_d\,\mathbf{g}\|^2}$$

  which is solved as a generalized eigenvalue problem. Distributed variants based on diffusion adaptation partition cost across nodes, significantly reducing computational load with only minor penalties in isolation performance (Zhang et al., 2023, Zhao et al., 2023).

• Joint Pressure and Velocity Matching (JPVM+):

  Boundary control via the Kirchhoff–Helmholtz equation yields improved broadband separation by matching both pressure and the radial component of particle velocity exclusively on zone contours, discarding the tangential velocity terms for enhanced robustness (Buerger et al., 2017).

• Perceptually Optimized Filters:

  Variable-span trade-off filters adjust between contrast and distortion by joint diagonalization and segment-wise masking based on psychoacoustic thresholds, offering up to +5 dB improved contrast and +20% perceptual quality/ intelligibility over standard methods (Lee et al., 2019).

3. Advanced Transduction: Parametric Loudspeaker Systems

Parametric array loudspeakers (PALs) utilize nonlinear demodulation of ultrasound carriers to create highly directional audible beams. The core modeling involves the Westervelt equation and Rayleigh integrals, with array pressure at an audio frequency generated as

$$p_a(r_m; \omega_a) = \rho_0 \omega_a / (4i) \int\!\!\int q(r_v) H_0^{(1)}(k_a |r_m - r_v|) d^2 r_v$$

PAL arrays substantially outperform conventional electro-dynamic loudspeaker arrays at frequencies above 4 kHz, showing up to 36.8 dB contrast at 8 kHz and retaining >28 dB contrast under severe amplitude perturbations (Zhuang et al., 2024). Recent multi-carrier designs enable virtual multi-channel output from a single ultrasonic transducer, dramatically collapsing hardware requirements while retaining mature array-processing algorithms (Zhuang et al., 24 Apr 2025).

4. Neural and Data-Driven PSZ Filter Synthesis

Deep learning frameworks, notably the Spatially Adaptive Neural Network (SANN-PSZ) and Binaural SANN (BSANN), parameterize listener position and head orientation via Fourier feature encoding and output multimodal filter coefficients for robust real-time zone rendering (Qiao et al., 2024, Jiang et al., 10 Jan 2026). Characteristic model specifications include:

• Input: normalized zone positions, frequencies, or target ATFs.
• Architecture: fully connected MLPs or 3D residual CNNs (for spatially distributed bright zones) (Zhu et al., 11 Dec 2025).
• Loss: composite terms accounting for bright zone accuracy, dark zone suppression, filter compactness, and explicit crosstalk cancellation.

Benchmark metrics from these models:

Method	IZI (dB)	IPI (dB)	XTC (dB)	NMSE (dB)	Data Compression
SANN-PSZ	12.8	11.5	—	−18.4	100×
PM	11.1	10.2	—	−16.7	baseline
BSANN (stereo)	10.23	11.11	10.55	—	10×

These neural frameworks yield strong isolation and computational efficiency (≥10× speedup), and can incorporate measured, simulated, and reverberant ATFs for enhanced robustness. BSANN, with integrated HRTF modeling and active XTC, achieves improved spatial fidelity and resilience to room asymmetries, closing IZI gaps and preserving interaural cues (Jiang et al., 10 Jan 2026).

5. Measurement, Metrics, and System Robustness

Performance assessment of PSZ systems utilizes:

• Inter-Zone Isolation (IZI):

  $$\text{IZI}_A = \min \left \{ \text{IZI}_A^{\text{corr}}, \text{IZI}_A^{\text{uncorr}} \right \}$$

  comparing energy ratios in predefined bright vs. dark zones.

• Inter-Program Isolation (IPI):

  Measures target program dominance over interference within a single zone.

• Acoustic Contrast (AC):

  The classical variant for single-channel systems.

Zone boundary maps reveal that PSZ shapes are complex and highly frequency-dependent, with mobility and positional shifts more severely degrading program isolation (IPI) than zone isolation (IZI) (Qiao et al., 2022). Robust filter selection with audio-based position tracking via normalized cosine similarity enables rapid convergence to near-optimal filters even under dynamic listener relocation (Bhattacharjee et al., 2024).

6. Specialized Techniques: Multipath Exploitation and Sound Sphere Systems

Private audio at specific spots is achievable by exploiting room multipath properties. By chunking messages and aligning filtered signals across multiple loudspeakers, constructive summation yields intelligible output only at designated spots; intelligibility metrics such as STOI reach 0.9 at targets and <0.2 elsewhere (Liu et al., 2018). In applications such as UltrasonicSpheres, amplitude-modulated ultrasonic carriers are broadcast by commodity speakers, with earphone-based DSP recovering user-selected streams and preserving spatial cues—all without external infrastructure or interference to non-wearers (Küttner et al., 3 Jun 2025). MCPL and PAL-based systems also bypass traditional multi-channel arrays via nonlinear demodulation strategies (Zhuang et al., 24 Apr 2025, Zhuang et al., 2024).

7. Practical Considerations and Future Directions

Effective PSZ design demands attention to:

• Zone sizing: limited by the center frequency; higher frequencies enable finer spatial separation, but at the cost of decreased zone area (Rafaely, 2023).
• Control point layout: sparse grids are feasible with neural models, but classical PM methods degrade under insufficient sampling (Zhu et al., 11 Dec 2025).
• Distributed computation: diffusion adaptation and ATC-LMS architectures scale efficiently with network size, reducing bottlenecks while maintaining performance (Zhao et al., 2023).
• Perceptual optimization: trade-offs between contrast and distortion can be balanced with variable-span filtering and psychoacoustic weighting (Lee et al., 2019).
• Reverberant and dynamic environments: robustness is critically dependent on training with realistic ATFs and rapid filter updates via head-tracking or position estimation (Qiao et al., 2024, Bhattacharjee et al., 2024).

Zone separation in time-varying or multi-user settings is likely to benefit from further algorithmic advances in adaptive learning, wide-band transduction, real-time DSP integration, individualized HRTFs, and comprehensive perceptual evaluation.

---

Personal Sound Zones constitute a rigorously defined, physically constrained, and increasingly versatile paradigm for spatial audio rendering, integrating active field control, advanced transducers, and deep neural synthesis to provide quantifiable sound isolation, perceptual quality, and configurational flexibility for shared environments and multi-user scenarios.