PROJECT 04 / SIGN LANGUAGE RECOGNITION

Teaching a computer to recognize hand signs.

This project explores how static hand signs can be cleaned, processed and matched to letters. The visual follows the journey from the original image to the final prediction.

STATIC HAND SIGN

STATIC LETTER SAMPLE

01 / ORIGINAL INPUT

02 / ISOLATED REGION

03 / PROCESSED MASK

04 / FEATURE RESPONSE

24 STATIC LETTER CLASSES

PREDICTED LETTER /

LSTATIC LETTER OUTPUT

This visual was created for the portfolio to show the project idea clearly.

01 / CONTEXT

What we were trying to solve

Images of hand signs contain background, lighting and shape differences that can make recognition difficult. The project explores how preprocessing can isolate the hand before classification.

The scope stays with static images rather than real-time webcam recognition.

02 / PROJECT SCOPE

A focused processing study

The project follows 24 static letter classes from image input to letter output.

It is an image-processing and classification experiment, not a finished accessibility product.

03 / PIPELINE

How the image moves through the system

Each stage removes a little more visual noise before the final classification.

  1. 01Original static image
  2. 02Isolated hand region
  3. 03Processed mask
  4. 04Extracted or learned features
  5. 05Letter classification
  6. 06Output

04 / PROCESSING VIEW

The image changes before the model sees it.

The supporting views compare the same L-shaped hand geometry through preprocessing, then show the static model flow.

PREPROCESSING / SAME HAND POSITION

ASL LETTER L

01 / ORIGINAL
02 / ISOLATED
03 / MASK
The index, thumb and folded fingers stay consistent across every stage.

STATIC-CLASS MODEL FLOW

24 STATIC LETTER CLASSES

  1. 01STATIC IMAGE
  2. 02PREPROCESSING
  3. 03FEATURES
  4. 04CLASSIFICATION
  5. 05LETTER OUTPUT
PREDICTED LETTERL
No percentage is attached to the conceptual feature response.

05 / REFLECTION

What I learned

The quality of the input can matter as much as the model that reads it.
  1. 01

    Preprocessing makes the hand shape easier to separate from its background.

  2. 02

    Comparing raw and processed images exposes where classification can become difficult.

  3. 03

    A technical pipeline is easier to understand when every stage remains visually consistent.

WHERE IT STANDS NOW

A static hand-sign recognition experiment and processing study.

The portfolio visual keeps the scope at 24 static letter classes and does not attach a confidence claim.

Got a role, a project or a good question? Say hello.

Email works best. LinkedIn works too. I am usually not far from either.