Portable device for presbyopia correction with optoelectronic lenses driven by pupil response

Presbyopia, a loss of accommodative range due largely to age-related stiffness of the crystalline lens and changes in the ciliary muscle, leads to difficulty focusing at near distances. Conventional corrections (progressive spectacles, multifocal contacts, intraocular lenses) involve trade-offs, invasiveness, or limited focus ranges. Prior opto-electronic solutions have explored tunable lenses and gaze-based control but often lack true mobility or energy-efficient integration. This work proposes Dynamic Auto-Accommodation Glasses: a fully wearable system that uses binocular pupil tracking on a smartphone to infer gaze distance and dynamically drive tunable lenses, aiming to restore a continuous focus range analogous to natural accommodation. The goal is to evaluate whether this mobile, energy-efficient, real-time system maintains or improves visual acuity across a broad range of viewing distances in presbyopic subjects.

Foundational studies (Fisher; Glasser & Campbell; Heys et al.) implicate increasing lens stiffness as a primary cause of presbyopia and document age-related declines in accommodation. Alternative presbyopia corrections include progressive or multifocal optics and surgical approaches, each with compromises in contrast, comfort, or flexibility. Recent research has proposed adaptive/tunable lenses and gaze-contingent systems including laser range finding or pupil tracking; however, many are benchtop systems or rely on non-mobile processing. Prior GPU-accelerated pupil tracking (e.g., Starburst algorithm implementations) has shown feasibility on desktops. This work advances the field by delivering a smartphone GPU-based binocular pupil tracking and control solution in a wearable device.

System: A fully portable prototype integrates two opto-electronic tunable lenses (Optotune EL-16-40-TC-VIS-20D) with drivers, two USB infrared cameras (UVC class), four IR LEDs, a 300 mAh 3.7 V battery for the LEDs, a 4-port micro-USB hub, and a Samsung Galaxy S7 smartphone (Exynos 8890, Android 7.0) as processor and controller. Cameras interface via UVCCamera library; lenses via COM port using UsbSerial. A custom 3D-printed frame houses components with a head pad for comfort. Calibration and control: A one-time calibration captures pupil positions while fixating far to establish reference centers and interpupillary distance. During operation, the phone continuously acquires frames (320×240 px) from both cameras; if processing of the prior frame is ongoing, the current frame may be dropped. A binocular pupil tracking algorithm (Starburst) implemented in OpenCL runs on the smartphone GPU. The system computes eye rotations and vergence-based gaze distance using trigonometric relations: estimates of pupil displacement on the ocular globe yield rotations α, β and, with interpupillary distance, compute target distance d via derived relations (A, B, γ). The inferred distance is mapped to required dioptric power, which is applied to the tunable lenses. Pupil tracking implementation: The Starburst algorithm includes preprocessing (Top-hat transform, thresholding, dilation) to locate corneal reflections and guide edge searches with radial rays detecting intensity transitions. Iterative center-of-mass convergence is followed by RANSAC ellipse fitting of pupil border points. OpenCL optimization on mobile GPU includes vectorized memory access, local memory, native math (native_sin, native_cos, native_sqrt), half-precision where feasible, kernel fusion (e.g., Top-hat + threshold), and separable filters. The system achieved near 24 fps processing of 320×240 video on the Galaxy S7. Response shaping: To mimic natural accommodation and avoid discomfort from abrupt power changes, lens power transitions are smoothed at 50 D/s. Accounting for camera capture/processing delays (<200 ms total response), intrinsic lens delay (~7 ms) and stabilization (~40 ms), the system delivers quick, smooth responses. Participants and visual acuity testing: Eight presbyopic subjects (mean age 62.5 years, range 52–78) underwent visual acuity measurements at six target distances (5 m, 1 m, 0.5 m, 0.3 m, 0.25 m, 0.2 m). Three conditions were tested: (1) far correction using trial lenses, (2) near correction using trial lenses (~2.5 D), and (3) Dynamic Auto-Accommodation Glasses. Visual acuity (decimal scale) was recorded at each distance. Processing performance was benchmarked using a 250-frame 320×240 video.

• Processing speed: The Galaxy S7 achieved nearly 24 frames per second on 320×240 input, enabling smooth real-time response to eye movements.
• Visual acuity across distances: With far correction, acuity declined as distance decreased; with near correction, best acuity was around 2–3.3 D; with the Dynamic Auto-Accommodation Glasses, acuity remained good across all tested distances (5 m to 0.2 m).
• Statistical comparisons: • At 0.5 m (2 D): far correction (M=0.59, SD=0.06) vs Auto-Accommodation Glasses (M=0.78, SD=0.14), paired t(7) = -4.3658, p = 0.003291. • At 5 m (0.2 D): near correction without glasses (M=0.27, SD=0.04) vs with glasses (M=0.80, SD=0.13), paired t(7) = -9.4305, p = 0.0000314. • At 0.2 m (5 D): near correction without glasses (M=0.37, SD=0.09) vs with glasses (M=0.83, SD=0.15), paired t(7) = -8.994, p = 0.00004284.

The Dynamic Auto-Accommodation Glasses achieve their aim of improving presbyopic visual acuity over a wide range of distances by using binocular pupil tracking to infer vergence and dynamically control tunable lenses. Compared to static far or near corrections, the device maintains near-constant acuity from 5 m to 0.2 m, demonstrating effective, continuous-focus behavior analogous to natural accommodation. The energy-efficient, smartphone GPU-based implementation provides real-time performance (~24 fps) and quick, smooth lens power transitions (<200 ms total, 50 D/s), addressing user comfort by avoiding abrupt changes. The broader tunable lens power range enables focusing at very close targets, surpassing typical adult accommodation. These results support the feasibility of a fully wearable, mobile, gaze-contingent presbyopia correction system for everyday use.

This work presents a fully portable, smartphone-powered Dynamic Auto-Accommodation Glasses that couple binocular pupil tracking with tunable lenses to restore a continuous focus range in presbyopic subjects. Visual acuity improved significantly across distances compared to static far or near corrections, and the mobile GPU implementation enabled real-time, smooth responses suitable for wearability. Future work should focus on improving ergonomics (reducing bulk and weight), extending battery life for full-day use, increasing field of view via larger-aperture lenses, and further enhancing robustness of control to minimize rare oscillations. Advances in opto-electronic lens technology and integrated hardware could facilitate broader adoption and performance gains.

• Field of view limited by lens aperture: approximately 43.6° per eye; binocular FOV ranges from ~53.1° at 0.3 m to ~43.9° at 10 m.
• Current prototype is somewhat bulky, reducing comfort for prolonged use.
• Battery capacity insufficient for 8 hours of continuous operation.
• Occasional control failures can produce abrupt power changes and oscillations; mitigated by improved control but still a consideration.
• Visual magnification changes during large dioptric transitions may require user adaptation.
• Limited lens aperture remains a technological challenge, though larger lenses may emerge in the future.