Skip to content

rohit-ptl/AIND-Recognizer

 
 

Repository files navigation

Artificial Intelligence Engineer Nanodegree

Probabilistic Models

Completed Project: Sign Language Recognition System

Introduction

The project implements hidden markov models for recognizing sign language from videos, given data on hand and nose positions. The solution meets the rubric here.

Take a look at the sample ASL recognizer video to see how the hand locations are tracked.

Summary of results

Features and Selectors
  1. Ground Features: Left and right hand x-y positions, relative to the nose position and normalized for each speaker.
  2. Polar Features: Ground features transformed to polar coordinates.
  3. Delta Features: One-period delta of ground features.
  4. Custom1 (C1): Difference in left and right hand locations (x and y)
  5. Custom2 (C2): Normalized C1

The selectors used were BIC (bayesian information criterion), DIC (discriminative information criterion), and CV (cross validated log likelihood).

Results

The scores from various combination of features and selectors were as follows (lower is better).

Selector BIC CV DIC Mean Min
Polar+Delta+C2 0.449438 0.477528 0.505618 0.477528 0.449438
Ground+Delta+C2 0.449438 0.511236 0.539326 0.500000 0.449438
Delta+C1+C2 0.455056 0.488764 0.516854 0.486891 0.455056
Polar+Delta+C1 0.466292 0.505618 0.516854 0.496255 0.466292
Ground+Polar+Delta 0.477528 0.595506 0.511236 0.528090 0.477528
Ground+Delta 0.483146 0.505618 0.500000 0.496255 0.483146
Ground+Polar+Delta+C2 0.483146 0.550562 0.522472 0.518727 0.483146
Ground+Delta+C1 0.488764 0.578652 0.522472 0.529963 0.488764
Ground+Polar+Delta+C1 0.505618 0.522472 0.567416 0.531835 0.505618
Ground+Polar+C2 0.528090 0.522472 0.505618 0.518727 0.505618
Ground+C2 0.556180 0.539326 0.550562 0.548689 0.539326
Ground+Polar 0.550562 0.567416 0.539326 0.552434 0.539326
Ground 0.550562 0.539326 0.573034 0.554307 0.539326
Ground+C1 0.539326 0.606742 0.539326 0.561798 0.539326
Polar 0.544944 0.561798 0.544944 0.550562 0.544944
Ground+C1+C2 0.567416 0.544944 0.567416 0.559925 0.544944
C1+C2 0.584270 0.623596 0.623596 0.610487 0.584270
C2 0.640449 0.601124 0.662921 0.634831 0.601124
Delta 0.612360 0.612360 0.623596 0.616105 0.612360
C1 0.629213 0.646067 0.674157 0.649813 0.629213
Mean 0.528090 0.555056 0.555337 0.546161

The best performing combination was Polar + Delta + C1 with a BIC selector.

About

Implemented Sign Language Recognition Project

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 86.9%
  • Python 13.1%