The problem FingerPulse-CV solves

Traditional human-computer interaction (HCI) and hardware automation rely on physical buttons, switches, or touch interfaces, which can be inefficient or inaccessible in hands-free, sterile, or industrial environments.

FingerPulse-CV solves this by providing a low-cost, real-time vision-based control interface:

• Touchless Hardware Control: Enables direct control of specific hardware pins (LEDs, relays, actuators) using simple hand gestures via a standard webcam.
• Granular Pin Mapping: Unlike generic finger-counting algorithms, FingerPulse-CV tracks individual fingers (Thumb, Index, Middle, Ring, Pinky) to trigger dedicated output channels independently.
• Accessible & Affordable: Works on basic microcontrollers (Arduino Uno/Nano) without requiring expensive sensor gloves or specialized hardware depth cameras.

Challenges I ran into

1. Serial Port Locking & Synchronization:

   • Issue: Connecting Python and Arduino over USB Serial frequently threw "Permission Denied" errors when the Arduino IDE Serial Monitor or leftover processes locked COM7.
   • Solution: Implemented explicit exception handling in Python using pyserial, ensuring port safety and graceful socket teardowns on application exit.

2. Thumb Tracking Accuracy:

   • Issue: Standard Y-axis landmark comparisons failed for thumb detection because the thumb moves laterally across the palm rather than folding down vertically like other fingers.
   • Solution: Switched to calculating Euclidean distance between thumb tip/IP landmarks and key palm anchors using math.hypot() for accurate lateral gesture recognition.

3. Data Overhead & Latency:

   • Issue: Continuously streaming serial data every frame caused micro-lag on the Arduino microcontroller.
   • Solution: Designed a state-checking system in Python that transmits encoded 5-bit binary payloads (e.g., 10000) only when finger states change.