Real-Time AUSLAN Fingerspelling Recognition Using Deep Learning (LSTM) | Australian Sign Language

Title: Real-Time AUSLAN Fingerspelling Recognition Using Deep Learning (LSTM) | Australian Sign Language Project

Description: Welcome to the demo of our Real-Time AUSLAN Fingerspelling Recognition System! This project leverages deep learning, specifically Long Short-Term Memory (LSTM) networks, to recognize fingerspelling in Australian Sign Language (AUSLAN) with high accuracy and real-time performance. 🎉

Project Motivation: While much research has focused on American Sign Language (ASL), AUSLAN remains underrepresented in gesture recognition technologies. This project seeks to bridge that gap, providing a robust and adaptable system tailored to AUSLAN, which can potentially be generalized to other sign languages.

Key Features and Challenges:

Hand Landmark Detection: We used MediaPipe to extract precise 3D landmarks of the hands, making our model efficient and effective.
Real-Time Accuracy: Our LSTM-based model achieved an impressive 99% accuracy, distinguishing subtle differences between gestures.

Challenges Solved:
Interclass Similarity: Using data augmentation (noise injection, scaling, and time warping) to help the model learn subtle differences between similar gestures.

Lighting Variations: Our landmark-based approach is resilient to different lighting conditions, ensuring reliable recognition.
Distance and Hand Size Variability: Keypoint normalization allows the model to be invariant to scale, adapting to various hand sizes and camera distances.

Why AUSLAN? 🇦🇺 As an often underrepresented sign language, AUSLAN deserves tailored technology that respects the nuances of its gestures. We hope this project will inspire further development in accessible technologies for the Australian Deaf community and beyond.

Demo Video Highlights: Explore how our model performs in real-time and handles various challenges.

Thanks for watching! If you found this project interesting, please like, share, and subscribe for more deep learning and AI-based solutions. Let’s make technology more inclusive! 🌏

#AUSLAN #DeepLearning #SignLanguageRecognition #AIForGood #MachineLearning #LSTM #ComputerVision #Inclusion #AustralianSignLanguage