Case Study · Accessibility & Computer Vision Research

PUGS

PUGS (Path Utility Guide System) is a camera-based assistive navigation research prototype built for blind and visually impaired users. The system evaluated whether a commonly available Android smartphone could provide environmental guidance without requiring specialized external sensors or navigation infrastructure.

As Researcher and Developer in a 2-student research team advised by faculty, I worked on the visual processing flow, obstacle proximity estimation, free path determination, and multimodal haptic/audio guidance.

Domain: Consumer (Accessibility & Assistive Tech)
My Role: Researcher + Developer
Team Context: 2 Student Researchers + Faculty Adviser
Year & Origin: 2022 · Thesis Research Project
Hardware Target: Android Smartphone (4 GB RAM Target)
Status: Completed Thesis Prototype

Overview & Research Focus

PUGS investigated the feasibility of utilizing standard Android smartphone camera hardware for real-time obstacle detection and spatial path guidance. Rather than relying on external GPS, wearable headsets, or specialized depth sensors, PUGS operated entirely on-device.

The research demonstrated that software-driven visual processing on accessible mobile hardware could assist visually impaired users in identifying walkable paths and avoiding near-field physical hazards.

Accessibility Goal Providing immediate environmental guidance without specialized external hardware (LiDAR, ultrasonic sensors, or beacons).

Engineering Focus Full-Stack · Architecture

Central Research Flow

The system narrative highlights how live camera input was processed into actionable tactile and acoustic navigation guidance:

Physical Environment
        ↓
Android Smartphone Camera Sensing
        ↓
On-Device Computer Vision Processing
        ↓
Detect General Objects / Physical Obstacles
        ↓
Estimate Object Location & Proximity Thresholds
        ↓
Identify Free / Walkable Path Regions
        ↓
Determine Suggested Direction (Left / Center / Right)
        ↓
Multimodal Haptic (Vibration) + Audio (Speech) Feedback
        ↓
User Adjusts Walking Path

Core Research Problem & Constraints

Visually impaired individuals face significant navigational challenges in unfamiliar indoor and outdoor environments. Dedicated assistive devices are often expensive, specialized, or require complex setup.

PUGS evaluated whether a mainstream Android smartphone (targeting devices with approximately 4 GB RAM or higher) could process environmental visual cues in real time to guide users safely.

Infrastructure Independence Operated without reliance on Google Maps, active GPS signals, Bluetooth beacons, or internet connectivity.

On-Device Architecture & Hardware Independence

PUGS used the smartphone itself as both the primary sensor and output terminal, avoiding external sensor attachments:

Input Sensing: Rear smartphone camera capture feeding continuous frame data to the processing loop.

On-Device Computation: Mobile visual frame processing for obstacle localization and free path detection.

Output Interaction: Haptic motor vibration patterns combined with Text-to-Speech (TTS) audio directions.

Spatial Frame Division & Path Selection

The application divided the camera frame into spatial zones to evaluate obstacle placement relative to the user's path:

Camera Frame Spatial Analysis:
┌─────────────────┬─────────────────┬─────────────────┐
│   LEFT ZONE     │   CENTER ZONE   │   RIGHT ZONE    │
│  Obstacle Check │  Obstacle Check │  Obstacle Check │
└─────────────────┴─────────────────┴─────────────────┘
        ↓                 ↓                 ↓
Proximity Check   Proximity Check   Proximity Check
        └─────────────────┼─────────────────┘
                          ↓
            Identify Free / Open Corridor
                          ↓
      Issue Directional Cue ("Steer Right / Path Clear")

Prototype Detection Capabilities

The PUGS research prototype was designed to detect and evaluate key environmental features:

General Obstacle Sensing: Identifying physical barriers in the immediate walking path.
Proximity Estimation: Evaluating obstacle distance thresholds (near vs. far warnings).
Free Path Identification: Locating unobstructed spatial corridors.
Directional Guidance: Suggesting safe direction adjustments (Left, Center, Right).
Class Recognition & Hazards: Recognizing specific object classes, stairways/steps, and edge drop hazards where supported by the vision prototype.

Restrained Research Claims The system functioned as an exploratory research prototype to demonstrate technical feasibility rather than guaranteed autonomous navigation.

Multimodal Haptic & Audio Feedback

HAPTIC VIBRATION SIGNALS

Pulsed vibration patterns for immediate obstacle proximity warnings
Directional haptic cues signaling left or right path adjustments
Silent tactile feedback useful in noisy public environments

AUDIO SPEECH GUIDANCE

Text-to-Speech verbal cues for object classification and warnings
Spoken directional cues ("Path clear ahead", "Obstacle left")
Clear acoustic feedback for blind user navigation

Research Validation & Prototype Testing

The research project evaluated prototype performance through controlled obstacle navigation trials and qualitative user testing feedback, gathering insights on detection latency, haptic response clarity, and walking path accuracy.

Verified Qualitative Outcomes

Feasibility Demonstration Validated that standard smartphone cameras and on-device processing can perform real-time assistive obstacle detection.

Hardware Independence Eliminated dependence on specialized wearable sensors, external depth cameras, or dedicated beacons.

Effective Multimodal Guidance Demonstrated that combining haptic vibration pulses with audio speech improves user directional awareness.

Academic Thesis Completion Successfully defended as an undergraduate Computer Science research thesis at Isabela State University.

Retrospective & Accessibility Engineering

Designing for Multimodal User Interfaces

PUGS provided early experience in designing accessibility-first user interfaces where non-visual channels (haptics and audio) serve as primary interaction mechanisms.

Mobile Performance Optimization

Processing live video frames on mobile devices highlighted the importance of algorithm efficiency, frame rate management, and memory optimization under resource constraints.